Partition Coefficient Measurements in Picoliter Drops Using a

Jan 14, 2009 - Nathan A. Marine,† Steven A. Klein,† and Jonathan D. Posner*,†,‡. Mechanical Engineering, and Chemical Engineering, Arizona Sta...
0 downloads 0 Views 244KB Size
Download PDF
Anal. Chem. 2009, 81, 1471–1476

Partition Coefficient Measurements in Picoliter Drops Using a Segmented Flow Microfluidic Device Nathan A. Marine,† Steven A. Klein,† and Jonathan D. Posner*,†,‡ Mechanical Engineering, and Chemical Engineering, Arizona State University, Tempe, Arizona 85287-6106 A microfluidic method to rapidly measure the octanolwater partition coefficient in thousands of individual picoliter drops is described. A T-junction microfluidic chip is used to generate a segmented flow of monodisperse, fluorescein-laden water in octanol carrier fluid. The partitioning of individual drops reaches equilibrium in less than 2 s. Epifluorescence microscopy is used measure the partition coefficient of fluorescein as a function of pH. Results compare well with previous measurements using traditional shake-flask methods. The methods presented here are rapid, provide detailed statistics, and can be run in parallel, enabling the simultaneous partitioning of thousands of compounds for various applications such as drug development, environmental testing, and combinatorial chemistry. Microfluidic partitioning and extraction in picoliter drops may be useful for studying molecules and particles away from their equilibrium state and in cases with limited samples. Partitioning of chemicals between water and organic phases is critical in determining their fate, accumulation, and toxicity in the environment and in the human body.1-3 The spatial and temporal mobility of a substance is largely determined by its solubility in water. The Environmental Protection Agency (EPA) uses partition coefficients to predict the environmental fate, aquatic toxicity, and bioaccumulation of chemicals and pollutants. Partitioning into an organic phase (typically octanol) is often used as a measure of bioavailability for herbicides, pesticides, pharmaceuticals, and potentially toxic chemicals. The octanol-water partition coefficient is a prerequisite for testing biological degradation and bioaccumulation in water for which the EPA Office of Prevention, Pesticides, and Toxic Substances (OPPTS) publishes harmonized test guidelines. Octanol-water partitioning is also an important physiochemical property of pharmaceutical drugs and agrochemicals and is a required measurement by legislation as part of the profile for high-volume production chemicals.4 The partition coefficient is measured by a variety of methods including * To whom correspondence should be addressed. E-mail: jonathan.posner@ asu.edu. Phone: 480-965-1799. † Mechanical Engineering. ‡ Chemical Engineering. (1) Turner, A.; Mawji, E. Environ. Sci. Technol. 2004, 38, 3081–3091. (2) Turner, A.; Mawji, E. Environ. Pollut. 2005, 135, 235–244. (3) Leo, A.; Hansch, C.; Elkins, D. Chem. Rev. 1971, 71, 525 ff. (4) Barzanti, C.; Evans, R.; Fouquet, J.; Gouzin, L.; Howarth, N. M.; Kean, G.; Levet, E.; Wang, D.; Wayemberg, E.; Yeboah, A. A.; Kraft, A. Tetrahedron Lett. 2007, 48, 3337–3341. 10.1021/ac801673w CCC: $40.75  2009 American Chemical Society Published on Web 01/14/2009

the shake (or stir) flask method,5,6 inverted high-performance liquidchromatography(HPLC),7-10 chromatographictechniques,11-14 titration techniques,15-17 and flow injection analysis (FIA) methods.18-20 One of the most common methods is some variant of the shake-flask method. The advantage of this method is that it is standardized, relatively simple, and is a direct measurement of the partition coefficient.11,18 The disadvantage of these methods is that they use large samples and typically require several days to complete. The time scale of these experiments is typically imposed by the mass transport of the sample across the relatively large distances in the flasks. Miniaturization of octanol-water partition coefficient analysis can offer rapid measurements with minute samples. FIA and variants of the monosegmented flow analysis (MSFA)21 have been usedwithgreataccuracytomeasureKow oforganiccompounds.18-20 These microvolume methods have reduced the measurement time down to as low as 4 min for a single measurement.21 In these methods, immiscible water and organic phases are loaded into a capillary such that the octanol and water phases share a large surface area for transport of the sample from one phase to the other. The large surface area and short distances result in rapid equilibrium of the sample. In FIA, the organic phase coats the walls of a PTFE tube, whereas in the MSFA the octanol and water form segmented plugs. Droplet-based (or segmented flow) microfluidic systems have been used to perform various processes such as chemical (5) Opperhuizen, A.; Serne, P.; Vandersteen, J. M. D. Environ. Sci. Technol. 1988, 22, 286–292. (6) Brooke, D.; Nielsen, I.; Debruijn, J.; Hermens, J. Chemosphere 1990, 21, 119–133. (7) Kaune, A.; Bruggemann, R.; Kettrup, A. J. Chromatogr., A 1998, 805, 119– 126. (8) Grouls, R. J. E.; Ackerman, E. W.; Korsten, H. H. M.; Hellebrekers, L. J.; Breimer, D. D. J. Chromatogr., B 1997, 694, 421–425. (9) Krop, H. B.; van Velzen, M. J. M.; Parsons, J. R.; Govers, H. A. J. Chemosphere 1997, 34, 107–119. (10) Terada, H. Quant. Struct.-Act. Relat. 1986, 5, 81–88. (11) Berthod, A.; Carda-Broch, S. J. Chromatogr., A 2004, 1037, 3–14. (12) Kleyle, R. M.; Nurok, D.; Kossoy, A. D.; Burris, S. C. J. Chromatogr., A 1996, 749, 211–217. (13) Dean, J. R.; Tomlinson, W. R.; Makovskaya, V.; Cumming, R.; Hetheridge, M.; Comber, M. Anal. Chem. 1996, 68, 130–133. (14) Marston, A.; Hostettmann, K. J. Chromatogr., A 1994, 658, 315–341. (15) Avdeef, A. J. Pharm. Sci. 1993, 82, 183–190. (16) Heerklotz, H. H.; Binder, H.; Epand, R. M. Biophys. J. 1999, 76, 2606– 2613. (17) Heerklotz, H.; Seelig, J. Biochim. Biophys. Acta 2000, 1508, 69–85. (18) Danielsson, L. G.; Zhang, Y. H. TrAC, Trends Anal. Chem. 1996, 15, 188– 196. (19) Gluck, S. J. Anal. Chim. Acta 1988, 214, 315–327. (20) Johansson, P. A.; Karlberg, B.; Thelander, S. Anal. Chim. Acta 1980, 114, 215–226. (21) Carlsson, K.; Karlberg, B. Anal. Chim. Acta 2000, 423, 137–144.

Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

1471

Figure 1. Schematic of the serpentine partitioning microdevice. Segmented flow is generated at the T-junction at the bottom left. Octanol and water flow from the west and north channels, respectively. The serpentine section is made up of 11 channels that are 10 mm long. The channels are 100 µm in width and 30 µm in height. The channels are fabricated using soft lithography of PDMS and are bonded to PDMS that is spin-coated onto a glass microscope slide.

synthesis and reactions,22,23 particle synthesis,24 extraction and purification,25 single-cell assays,26 creation of emulsions,27-29 and protein crystallization,30-32 among others.33 In this paper, we present a segmented flow microfluidic method for measuring octanol-water partitioning coefficients in single picoliter drops. Picoliter water droplets are generated in octanol carrier fluid within a T-shaped segmented flow device fabricated in poly(dimethylsiloxane) (PDMS). The partition coefficient of fluorescein, a fluorescent dye, is measured as a function of pH using epifluorescence microscopy.The microfluidic partitioning measurements reach equilibrium in seconds for a single drop and are conducted in minutes for thousands of individual picoliter droplets. The methods presented here are rapid, provide detailed statistics, and can be run in parallel enabling the simultaneous partitioning of thousands of compounds for various applications such as drug development, environmental testing, and combinatorial chemistry. EXPERIMENTAL SETUP AND METHOD Microchip Design. Partitioning microdevices with serpentine channels and a T-shaped segmented flow injector are shown in (22) Sarrazin, F.; Prat, L.; Di Miceli, N.; Cristobal, G.; Link, D. R.; Weitz, D. A. Chem. Eng. Sci. 2007, 62, 1042–1048. (23) 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, 484–487. (24) Millman, J. R.; Bhatt, K. H.; Prevo, B. G.; Velev, O. D. Nat. Mater. 2005, 4, 98–102. (25) Mary, P.; Studer, V.; Tabeling, P. Anal. Chem. 2008, 80, 2680–2687. (26) He, M. Y.; Edgar, J. S.; Jeffries, G. D. M.; Lorenz, R. M.; Shelby, J. P.; Chiu, D. T. Anal. Chem. 2005, 77, 1539–1544. (27) Sugiura, S.; Nakajima, M.; Iwamoto, S.; Seki, M. Langmuir 2001, 17, 5562– 5566. (28) Utada, A. S.; Lorenceau, E.; Link, D. R.; Kaplan, P. D.; Stone, H. A.; Weitz, D. A. Science 2005, 308, 537–541. (29) Anna, S. L.; Bontoux, N.; Stone, H. A. Appl. Phys. Lett. 2003, 82, 364–366. (30) Hansen, C. L.; Sommer, M. O. A.; Quake, S. R. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 14431–14436. (31) Hansen, C. L.; Skordalakes, E.; Berger, J. M.; Quake, S. R. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 16531–16536. (32) Hansen, C. L.; Classen, S.; Berger, J. M.; Quake, S. R. J. Am. Chem. Soc. 2006, 128, 3142–3143. (33) Song, H.; Chen, D. L.; Ismagilov, R. F. Angew. Chem., Int. Ed. 2006, 45, 7336–7356.

1472

Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

Figure 1. Segmented flow of octanol and water is generated at the T-intersection at the bottom left of Figure 1. The octanol and water, respectively, flow from the west and north channels of the T-intersection. The segmented drops of octanol and water flow through the serpentine channels to the waste port at the top right of Figure 1. The injection channels that form the T are 3.4 mm long. Each serpentine channel is 10 mm long. The channels are 100 µm wide and 30 µm tall. Microfabrication. The devices were fabricated using soft lithography of PDMS.34 We use an SU8 (2025 MicroChem. Corp., Newton, MA) master template fabricated on a Si(100) wafer (University Wafer Corp., Boston, MA) using photolithography. Sylgard 184 PDMS prepolymer (Dow Corning, Midland, MI) at 30:1 base polymer/curing agent (A/B) is then cast on a silanized master. The PDMS is then cured at 80 °C in a convection oven for 30 min. The cured PDMS is peeled off from the master and bonded by baking for 1 h at 80 °C onto PDMS 3:1 (A/B) that is spin-coated onto a glass microscope slide. The bond strength between the two layers of PDMS was increased by using dissimilar monomer/hardener ratios.35,36 Experimental Setup. We used quantitative epifluorescence microscopy on an upright microscope (AZ 100, Nikon, Melville, NY) to image the flows. Images were recorded on a cooled CCD camera (Cascade IIb, Photometrics, Tucson, AZ) with a blue-green filter cube (excitation 450-490 nm, emission 510-570 nm, XF1002, Omega Optical, Brattleboro, VT). The images are 20 × 20 pixels and are recorded at ∼30 frames/s with a 40× objective with a numerical aperture of 0.5. We used two independent (KD Scientific model 210, Holliston, MA) syringe pumps to provide constant flow rate of the octanol and water. The octanol and water flow rates (Qoct and Qwater, respectively) were set between 0.3 and 1.0 µL/min for all measurements which results in a Peclet number range of 210-680. The ratio of Qwater/Qoct was typically unity so that the length of the droplets was equal.37 Glass, 50 µL syringes (Fisher Scientific, Waltham, MA) and 23 gauge needles (0.5 in. long, type 304, i.d. 0.017 in., o.d. 0.025 in.) were used. Fluidic connections were made using Tygon tubing (1/16 in. i.d., McMaster-Carr, Santa Fe Springs, CA) and stainless steel tubes (NE-1300-01, New England Small Tube Corp., Litchfield NH). Chemicals. We measure the partitioning of fluorescein disodium salt (Fisher Scientific, Waltham, MA) in HEPES buffered aqueous solutions and 1-octanol (CAS no. 111-87-5 Acros Organics, Geel, Belgium). The partition coefficient is measured over a range of pH 6.4-8.3 using 100 mM HEPES buffer (Sigma-Aldrich, St. Louis, MO). The buffer pH is controlled by the concentration of the sodium phosphate monobasic and sodium phosphate dibasic heptahydrate in the solution. These conditions are chosen consistent with previous partitioning measurements of fluorescein of Grimes et al.38 Although surfactants are typically used in (34) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974–4984. (35) Unger, M. A.; Chou, H. P.; Thorsen, T.; Scherer, A.; Quake, S. R. Science 2000, 288, 113–116. (36) Liu, M.; Chen, Q. F. J. Micro/Nanolithogr. MEMS MOEMS 2007, 6, 023008. (37) Garstecki, P.; Fuerstman, M. J.; Stone, H. A.; Whitesides, G. M. Lab Chip 2006, 6, 437–446. (38) Grimes, P. A.; Stone, R. A.; Laties, A. M.; Li, W. Arch. Ophthalmol. 1982, 100, 635–639.

segmented flow devices,39 here we do not use them to avoid any interference with the partitioning. THEORY The partition coefficient is defined as Kow )

Co Cw

(1)

where C is the concentration of fluorescein and the subscripts o and w denote octanol and water, respectively.40 Here we are measuring the partitioning of fluorescein which is an ionizable compound. Partitioning of ionizable compounds is often described by distribution coefficients, apparent partition coefficients, or partitioning ratios.40 We consider our experiments to be direct measures of the partitioning ratio, following the previous fluorescein partitioning measurements of Grimes et al.38 and the definitions set forth by Leo et al.3 For the range of concentrations used in our experiments, the fluorescence intensity is linearly related to fluorescein concentration. Assuming negligible partitioning into the PDMS and equal and linear fluorescence response in both solvents, one can directly measure the partition coefficient from fluorescence using Kow ) Io/Iw, where I is the fluorescence intensity in the (o) octanol and (w) water drops. We found that fluorescein fluorescence in octanol is red-shifted 70 nm such that the fluorescence captured using the epifluorescence filters is rather weak. Two-dimensional fluorescence spectra displaying the spectral red shift in octanol are shown in the Supporting Information. Due to the weak fluorescence of fluorescein in octanol, we infer the sample concentration in octanol from the concentration in the water sample. The number of moles in the octanol phase can be written as no ) nw,initial - nw,final. Here the subscripts final and initial denote the number of moles of fluorescein, respectively, in a water drop at equilibrium and in the initial state. The molar concentration in the water and octanol are simply Cw ) nw/Vw and Co ) no/Vo. The concentration of a dilute sample in water can be measured as Cw ) βIw, where β is constant. Therefore, the number of moles in a water drop can be written as nw ) VwIwβ. Rewriting eq 1 in terms of intensity of the water drop we get Kow )

[

]

Iw,initial Vw -1 Iw,final Voct

(2)

Since the channel depth is uniform throughout the channel, we can approximate the volume ratio of the drops as Vw/Vo ) Lw/Lo, where L is the length of the droplet in the microchannel. In our experiments we infer the length ratio of the drops from the intensity-time plots as shown in Figure 4 and described in the Results and Discussion section. Correcting for the background signal, the partition coefficient in terms of measured variables is Kow )

[

]

Lw Iw,initial - Idark -1 Lo Iw,final - Idark

(3)

(39) Thorsen, T.; Roberts, R. W.; Arnold, F. H.; Quake, S. R. Phys. Rev. Lett. 2001, 86, 4163–4166. (40) Sangster, J. Octanol-Water Partition Coefficients: Fundamentals and Physical Chemistry; John Wiley & Sons: West Sussex, U.K., 1997.

where the subscript dark denotes the fluorescence measured by the camera in the channel with no fluorescein. The fluorescence intensity of each droplet is determined by area averaging over the droplet area and then ensemble averaging over the multiple frames that constitute a single droplet. Equation 3 assumes that no fluorescence is lost due to photobleaching and that no mass is lost to the channel walls. To avoid photobleaching of the initial intensity images the light source remains blocked for 2 min before taking data. For the final intensities, we capture images of the droplets by taking 50,000 continuous frames at the end of the serpentine channel which takes approximately 15 min. Doing this requires the light source to be illuminating a fraction of the chip the entire measurement period. We tested for photobleaching under these conditions but found that there was no measurable photobleaching because of the low light intensity used and the limited time each droplet is exposed to the light. Equation 3 also assumes no mass is lost to the channel walls. However, the network polymer structure of PDMS is well-known to be permeable to and absorb water and solvents,41 gases,42,43 and small molecules.44 PDMS has been explored for solid-phase microextraction of organic compounds from aqueous solutions, and the equilibrium partitioning of compounds in PDMS-water have been related to octanol-water partitioning coefficients.45-48 For these reasons we have run several experiments to determine the effects of partitioning into the PDMS microstructure. We filled the PDMS channels with fluorescein dyes for 24 h. As we expected, due to fluorescein’s relatively polar nature, we found that there was no partitioning of fluorescein into PDMS. However, for an apolar molecule like rhodamine, we and others have observed significant partitioning.44 Partitioning of the apolar sample molecules into the PDMS microchip may result in a loss of mass depending on the time scale of the experiment and the channel surface to volume ratio. We measured the diffusivity of rhodamine dye in PDMS and found it to be 1 × 10-11 m2/s, which is an order of magnitude less than in water. See the Supporting Information for details of the experiment and results. This result suggests that if the experiments can be conducted quickly that the sample molecules may reach equilibrium in the water and octanol phases before significant mass is lost to the microchannel walls. For example, here fluorescein dye reaches equilibrium partitioning in less than 2 s which would translate to a penetration length in PDMS of only 6 µm. The microchannel material, sample molecule polarity, and experiment time scale must all be considered to ensure accurate quantitative measurements. In this case, fluorescein is relatively polar and does not absorb into the PDMS by any measurable amount. Alternatively, to avoid possible partitioning into the microfluidic substrate, the channels can be fabricated in inert materials such as glass or silicon. (41) Lee, J. N.; Park, C.; Whitesides, G. M. Anal. Chem. 2003, 75, 6544–6554. (42) Singh, A.; Freeman, B. D.; Pinnau, I. J. Polym. Sci., Part B: Polym. Phys. 1998, 36, 289–301. (43) McDonald, J. C.; Whitesides, G. M. Acc. Chem. Res. 2002, 35, 491–499. (44) Toepke, M. W.; Beebe, D. J. Lab Chip 2006, 6, 1484–1486. (45) Beltran, J.; Lopez, F. J.; Cepria, O.; Hernandez, F. J. Chromatogr., A 1998, 808, 257–263. (46) DeBruin, L. S.; Josephy, P. D.; Pawliszyn, J. B. Anal. Chem. 1998, 70, 1986–1992. (47) Dugay, J.; Miege, C.; Hennion, M. C. J. Chromatogr., A 1998, 795, 27–42. (48) Baltussen, E.; Sandra, P.; David, F.; Janssen, H. G.; Cramers, C. Anal. Chem. 1999, 71, 5213–5216.

Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

1473

Figure 2. (a) Fluorescence micrograph of fluorescein partitioning in a segmented flow of water and octanol (0.5× objective). The fluorescein partitions from the monodisperse, picoliter water droplets (bright) into continuous octanol phase (dark). The water droplets become darker as they travel along the channel because fluorescein partitions into the octanol where it does not fluoresce as brightly under the same excitation wavelength in octanol. Segmented flow is generated by injecting fluorescein in HEPES buffer (north channel on left) into a continuous stream of octanol (bottom left). (b) Segmented flow created at the T-junction of PDMS microchannel (10× objective) showing droplet breakup and capillary waves.

Figure 3. Normalized intensity of fluorescein in a water droplet as a function of time. The intensity is normalized by the intensity in an unpartitioned droplet. Droplet intensity decays exponentially until reaching equilibrium. The exponential decay time constant is 0.37, and the time to reach equilibrium is approximately 1.2 s.

RESULTS AND DISCUSSION Figure 2a shows an image of the partitioning device with segmented flow of fluorescein in buffered aqueous water in a carrier fluid of octanol. This image was recorded with a 0.5× magnification. The segmented flow is generated at the Tintersection (lower left) by flowing 1-octanol from the west channel well and fluorescein-laden HEPES buffer from the north well (of the T-intersection) as shown in Figure 2a. The droplets become darker over time as they travel down the channel and the fluorescein partitions into the octanol. However, the octanol phase remains dark throughout the channel because fluorescein in octanol has relatively weak fluorescence in spectral range of the epifluorescence filters used here. Figure 2a shows that the intensity of the water droplets looks nearly equilibrated by the fourth drop, midway through the first channel. Figure 3 shows the normalized water droplet fluorescence as a function of time. The time axis in this plot is a time in the Lagrangian reference frame from the point of creation and can be equated to an Eulerian measure of distance using the average velocity in the channel. The dye diffuses and partitions into the octanol phase resulting in an exponential decay of the fluorescence. The exponential decay time constant is 0.37, and the time to reach equilibrium is 1474

Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

Figure 4. Normalized, area-averaged intensity of the droplets vs time measured at the outlet of the microchannel. The large values are the water droplets, and the low values are the octanol phase between the droplets. The intensity is normalized over the initial (unpartitioned) intensity of the fluorescein. The upward and downward triangles denote the front and back of the water drops, respectively. The partition coefficient of individual drops are determined by ensemble averaging all frames of a single droplet (points between the triangles).

approximately 1.2 s, which is consistent with Figure 2a. This equilibrium time agrees well with the predictions established by Mary et al. based on the diffusive time scale (L2/D) and the Peclet number using the droplet length as the characteristic length scale over which mass transport occurs.25 For our system the time required to reach equilibrium is small relative to the time required to travel through the channel (less than 2 s compared to 1 min). Figure 2b shows a magnified view of the T-intersection where the droplets break up. For our system the capillary number Ca ) µu/γ is ∼2 × 10-3, where µ is the fluid viscosity, u is the fluid velocity, and γ is the surface tension. This value is less than the critical value of 10-2 established by Garstecki et al.; therefore, the breakup of the droplets is controlled by the Laplace pressure drop across the droplets as they form.37 The small capillary waves on the surface of the water drops in Figure 2b are evidence of surface tension dominated behavior associated with small capillary number flows. The length of the droplets can be controlled by adjusting the ratio Qwater/Qoct.37 Controlling the length of the droplets is important to partitioning and extraction as it dictates the length scale over which over which mass transport occurs. The lengths of the drops are approximately 300 µm, resulting in drop volume of ∼900 pL. Figure 4 shows a plot of the normalized intensity of several individual picoliter drops. The regions of the plots with large intensities are the water drops, whereas the low values correspond to the octanol phase. This data is obtained by monitoring the fluorescence near the outlet of the microchannel (top right in Figure 2a). Each data point in Figure 4 is an area average of the initial, final, and background fluorescence over 20 × 20 pixels. We measure the initial and background intensities at the beginning of each experiment. Each data point in Figure 4 is determined from a single frame of a single droplet. We use the data shown in Figure 4 to obtain quantitative measurements of the partition coefficient. When calculating Kow for each drop we ensemble average over all of the frames containing a single droplet. The beginning and end of drops are, respectively, denoted as upward and downward pointing

Figure 5. Histogram of the normalized partition coefficient at pH 7.93 of nearly 1000 individual 900 pL drops. The x-axis unit represents one standard deviation, and the graph is centered about the mean.

triangles. The drops with intermediate intensities are the interface of the droplets and are not considered when calculating the average intensity. The length of each droplet is also determined using the same criteria, where the intermediate intensities are shared equally between the water and octanolwater drops when calculating the length ratio of eq 3. Figure 5 shows the histogram of the normalized partition coefficient ¯ ow)/σK at pH 7.93 of nearly 1000 individual drops. (Kow - K ow The bar denotes an ensemble average over many drops, and σ is the standard deviation. Thus, each x-axis unit represents one standard deviation and the graph is centered about the mean. The mean partition coefficient of this data set is 0.18, the standard deviation is 0.08, and the coefficient of variance (CV) is 0.44. In order to validate the microfluidic segmented flow methods used here, we map the partition coefficient over a range of pH values pH 6-8 used by Grimes et al. Figure 6 shows a plot of the measured Kow at pH 6.43, 6.77, 7.12, 7.66, 7.80, 7.93, and 8.30. The upright triangles show the measurements of Grimes et al., whereas the squares show the average measurements made with the microfluidic method. Each dot represents the partition coefficient of a single drop, and the error bars are two standard deviations. For pH 7.66, 7.8, 7.93, and 8.26 the error bars are not visible because 2σ < 0.3. The figure shows that the average values obtained with the microfluidic method correspond very well with the trend established by Grimes et al. At pH 6 the fluorescein partitions highly into the octanol phase, while starting at pH 7.4, the amount of partitioning levels off almost to zero. In our measurements we see that when a molecule partitions highly into the octanol phase that there is a wide variability in the partition coefficient. Despite the large variability, the CV is similar in all of the measurements indiscriminate of pH. The variance is more apparent at high partitioning coefficient (low pH) because both the standard deviation and the mean are larger. In fact, the CV is actually larger in the high pH (0.49) than the low pH (0.36) although the variance appears larger at low pH values. A complete uncertainty analysis is presented in the Supporting Information. By examining the individual components in the uncertainty we see that the uncertainty in the intensity is negligible, whereas the main source of variability arises from the uncertainty in determining the length of the droplets. We expect the variability in the intensity to be small since the fluorescent detection method for fluorescein is very sensitive. The sensitivity of the fluorescence

Figure 6. Octanol-water partition coefficient of fluorescein as a function of the buffered aqueous phase pH. The upright triangles, 4, are shake-flask measurements obtained from Grimes et al. (ref 38). The open squares, 0, are our average measurements calculated from nearly 1000 drops each. Each dot represents the partition coefficient of a single drop, and the error bars are two standard deviations. For pH 7.66, 7.8, 7.93, and 8.26 the error bars are not visible because 2σ < 0.3.

detection method depends on a number of factors including the sensitivity of the camera, the numerical aperture of the imaging system, the octanol-water volume ratios used, the intensity of the light source, the depth of the microchannel, and the concentration and fluorescence cross section of the sample being analyzed. Since the variability from the intensity is negligible, the uncertainty is not a function of pH. Rather, the variability is controlled by the length of the droplets in a particular experiment and is largest when the droplets are small. Therefore, we can minimize the variability of our measurements by creating longer droplets or decreasing the uncertainty in determining the droplet length. CONCLUSION We have presented a microfluidic method for rapid measurement octanol-water partition coefficient in thousands of picoliter drops. Picoliter water droplets are generated in octanol carrier fluid within a T-shaped segmented flow device fabricated in PDMS. Quantitative measurements of fluorescein partition coefficient are measured as a function of pH which compare well with published values. Here, fluorescein concentration is measured using epifluorescence microscopy. This segmented flow partitioning method may also be applicable to a wide range of nonfluorescent compounds using alternative on-chip detection methods such as UV-vis absorption. UV-vis detection has already been applied to segmented flows in capillaries21 and to on-chip single-phase flows.41,49-54 Alternatively, these methods can be adapted into hyphenated platforms with external analytical equipment where the segmented flow output is phase-separated to obtain separate water and octanol streams55 and then analyzed using existing (49) Petersen, N. J.; Mogensen, K. B.; Kutter, J. P. Electrophoresis 2002, 23, 3528–3536. (50) Billot, L.; Plecis, A.; Chen, Y. Microelectron. Eng. 2008, 85, 1269–1271. (51) Salimi-Moosavi, H.; Jiang, Y. T.; Lester, L.; McKinnon, G.; Harrison, D. J. Electrophoresis 2000, 21, 1291–1299. (52) Jackman, R. J.; Floyd, T. M.; Ghodssi, R.; Schmidt, M. A.; Jensen, K. F. J. Micromech. Microeng. 2001, 11, 263–269. (53) Vlckova, M.; Kalman, F.; Schwarz, M. A. J. Chromatogr., A 2008, 1181, 145–152. (54) Gustafsson, O.; Mogensen, K. B.; Ohlsson, P. D.; Liu, Y.; Jacobson, S. C.; Kutter, J. P. J. Micromech. Microeng. 2008, 18, 055021. (55) Gunther, A.; Jhunjhunwala, M.; Thalmann, M.; Schmidt, M. A.; Jensen, K. F. Langmuir 2005, 21, 1547–1555.

Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

1475

analytical techniques amenable to small samples (e.g., mass spectrometry). The advantage of our microfluidic method is that partitioning within individual picoliter drops reaches equilibrium within seconds while large statistics across thousands of drops can be conducted in minutes. The microfluidic method presented here is rapid, provides detailed statistics, requires small samples, and can be run in parallel enabling the simultaneous partitioning of thousands of compounds for various applications including partitioning or extraction of compounds far from equilibrium where bulk methods may not be appropriate. ACKNOWLEDGMENT N.M. appreciates Philip M. Wheat and Carrie Sinclair for their help with microfabrication of PDMS devices. He also acknowledges Kiril Hristovski and Paul Westerhoff for help in obtaining the fluorescent spectra data for fluorescein. This work was

1476

Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

supported by DOE Grant No. DE-FG02-08ER64613 with Daniel Drell as grant monitor. The authors acknowledge the use of facilities within the Center for Solid-State Electronics Research at Arizona State University. SUPPORTING INFORMATION AVAILABLE Additional information regarding a detailed uncertainty analysis, the partitioning of the sample in the PDMS substrate, and the fluorescence spectra of fluorescein in water and octanol. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review August 8, 2008. Accepted December 13, 2008. AC801673W