Letter pubs.acs.org/ac
Cite This: Anal. Chem. XXXX, XXX, XXX−XXX
Automated Femtoliter Droplet-Based Determination of Oil−Water Partition Coefficient Miaosi Li,*,† Brendan Dyett,‡ and Xuehua Zhang*,§ †
School of Engineering, RMIT University, Melbourne, Victoria 3000, Australia School of Science, RMIT University, Melbourne, Victoria 3000, Australia § Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta T6G 1H9, Canada ‡
Downloaded via KEAN UNIV on July 20, 2019 at 16:25:10 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
S Supporting Information *
ABSTRACT: The oil−water partition coefficient of organic compounds is an essential parameter for the determination of their behaviors in environments, food, drug delivery, and biomedical systems, just to name a few. In this work, we establish a highly efficient approach to quantify the partition/distribution coefficient using surface femtoliter droplets. In our approach, droplets of 1-octanol were produced on the surface of a solid substrate in contact with the flow of an aqueous solution of the analyte. The analyte was rapidly enriched in the droplets from the flow and reached the partition equilibrium in a few seconds. The entire procedure was automated by continuous solvent exchange, and the analyte partition in the droplets was quantified from the in situ UV−vis spectrum collected by a microspectrophotometer. Our approach was validated for several substances with the octanol−water partition/distribution coefficient ranging from −1.5 to 4, where our results were in good agreement with the values reported in the literature. This method took ∼3 min to detect one analyte with the volume of the organic solvent at ∼50 μL. Thus, our surface droplet platform can greatly minimize the consumption of both solvent and analytes and can shorten the time for the determination of the partition of new compounds, which overcomes the drawbacks of the traditional shake-flask method and presents excellent reproducibility, high accuracy, cost-effectiveness, and labor-saving operation. The highly efficient micro/nanoextraction, partition, and real-time detection enabled by the surface droplets has the potential for many other high-throughput applications.
U
is advanced in its simple operation, low cost, and accuracy. The major drawback of the shake-flask method, however, is that it is time-consuming. Due to the limited interface between the two immiscible phases, it requires more than 24−48 h for a chemical to reach the partition equilibrium in the 2 phases. In addition, this method requires a large quantity of the solvents, which raises the cost of screening expensive drugs and leads to environmental concerns from the waste solvents. Although researchers can now estimate the coefficient through the molecular fragment-based calculation,11,12 the low accuracy of this method limits its use in practical applications.7 A micro- or nanoscale droplet-based platform holds the most promising potential for the miniaturization of the screening of partition coefficients to overcome the disadvantages of the shake-flask method. A few reported works have examined the partition coefficient by microdroplets produced in the bulk solution or in microfluidics technologies.13,14 The oil-like droplets suspended in the water phase maximize the surface to enhance the partition efficiency of the substance while minimizing the volume of the organic solvent.15−17 However, the droplets still need to be collected, and the online detection
nderstanding the partition behaviors of drug and pollutant molecules between water and organic phases is critical in the assessment of their environmental fate, bioaccumulation factors, and organism toxicity.1,2 A parameter, Ko‑w, usually expressed in a logarithm form, log Ko‑w, is defined as the partition coefficient referring to the equilibrium concentration ratio of a chemical in the organic phase (normally n-octanol) and the aqueous phase. Hydrophobic substances of high log Ko‑w values in octanol−water systems possess high absorbing ability in soils or sediments and tend to accumulate in living organisms. In contrast, hydrophilic chemicals will primarily remain in aqueous regions or blood serum.3−5 On the basis of these effects, the octanol−water partition coefficient has now become one of the standard physical properties in the registration of any new drug or organic compound. The most commonly used method to measure the partition coefficient is the “shake-flask” method performed in a separation tunnel, where the partition equilibrium of the compound is attained in the mixture of octanol and water phase.1,6 The concentration of the compound in each phase is measured upon the separation of the two immiscible phases. Compared to other ways such as inverted high-performance liquid chromatography (HPLC),7 flow injection analysis (FIA),8,9 or the slow-stirring method,10 the shake-flask method © XXXX American Chemical Society
Received: June 5, 2019 Accepted: July 11, 2019 Published: July 11, 2019 A
DOI: 10.1021/acs.analchem.9b02586 Anal. Chem. XXXX, XXX, XXX−XXX
Letter
Analytical Chemistry
Figure 1. (a) Scheme of the automated process showing surface droplet formation, partition detection, and surface renewing. (b) Size distribution analysis of the octanol droplets and the optical microscopic image showing morphology of the droplets (inlet). (c) Time series images showing the partition process of the R6G in aqueous solution in contact with octanol droplets. (d) The measured intensity of R6G in droplets against the partitioning time. (e) The intensity of the droplets at the end of the partition process for five cycles. Scale bar: 50 μm.
detection cycles. The lateral diameter of the droplets ranges from 10 to 50 μm with an average of ∼30 μm. There is no noticeable deviation in the size of the droplets within five cycles, suggesting a well-controlled droplet formation through solvent exchange. The droplets may contain octanol−ethanol− water immediately after their formation; however, by the end of the solvent exchange, ethanol in the droplets is removed by the flow of water, and hence, the droplets consist of octanol in equilibrium with water. The partition of the analyte between droplets and water is comparable to that of the bulk mixing in the shake-flask method. We chose a common fluorescent dye, rhodamine 6G (R6G), as our model compound to reveal the dynamical features of our droplet-based method. The reported partition coefficient, log Ko‑w, of R6G in octanol−water is 2.69 ± 0.18,23 suggesting a higher concentration of R6G in octanol phase than in the water phase. We conducted solvent exchange at a slow flow rate of the R6G solution at 200 μL/min (velocity 3.2 mm/s, see the Supporting Information) to follow the progressive distribution of R6G into octanol droplets, as shown in Figure 1c. The fluorescence intensity of R6G in droplets increases as a function of the flow time in Figure 1d and approaches a plateau at ∼80 s, demonstrating that the partition of R6G in water and in droplets reached the equilibrium. If required, the flow rate can be further increased, and the time for the partition equilibrium can be shortened to a few seconds. We measured the maximum fluorescence intensity of the droplets for five repeated cycles, as shown in Figure 1e. The fluorescence image at 80 s of the flow from the fifth cycle is also added in Figure 1c as comparison to the one from the first cycle. The consistency of the droplet morphology and the intensity of R6G in the droplets for each detection cycle clearly shows the great reproducibility of this approach in conducting continuous screening of multiple compounds. The droplets in the flow channel were examined by a microspectrophotometer, which is simply a microscope combined with an ultraviolet−visible (UV−vis) spectrometer. The spatial resolution of the instrument allows for the real time detection of the R6G concentrations in octanol droplets and in the surrounding water solution (see equipment information
for the analytes, especially for nonfluorescent molecules, still remains a challenge.13,14 Surface femtoliter droplets formed at liquid−solid interfaces exhibit unique and advanced features as an alternative platform to determine the partition coefficient.18−20 In particular, (1) the production of surface droplets by solvent exchange is inherently scalable and ultrafast, i.e., 106 droplets per second;21 (2) the large surface-to-volume ratio of femtoliter droplets provides extremely efficient liquid−liquid microextraction to the compounds in aqueous solution, as so-called “nanoextraction”;20,21 (3) these oil droplets are located on a solid surface in contact with an immiscible aqueous medium, which does not require an additional step for separation or collection of droplets from the bulk liquid mixture and allows for online detection of the extracted compounds;22 and (4) the droplets are stable during continuous liquid flow through the channel.18 All mentioned features of the surface droplets are desirable for automated and high throughput screening of the oil−water partition coefficient. Here, we developed a novel and highly efficient approach to quantify the partition coefficient for drug and pollutant molecules based on surface femtoliter droplets. The process is fully automated by programming the sequence of the liquids that flow through a channel where the droplets are located on the wall. Our approach is general, versatile, and applicable for continuous detection of multiple compounds. The entire process takes ∼3 min and requires no more than 50 μL of organic solvent for the detection. The results were supported by reported values, which demonstrates the accuracy and reliability of this droplet-based method. The process is illustrated in Figure 1a. The details of assembling the flow chamber and the syringe pump were reported in our previous work.19 Here, the sequence of the liquid injecting into the flow chamber is programmed (see details in the Supporting Information). As such, the whole procedure, including droplet formation, liquid−liquid extraction, online detection, and droplet removal, is automated in a single chamber. Figure 1b shows a representative image of surface droplets (inlet) and the size distribution of droplets performed in five B
DOI: 10.1021/acs.analchem.9b02586 Anal. Chem. XXXX, XXX, XXX−XXX
Letter
Analytical Chemistry
solution with known concentrations in the flow channel on the bare substrate. As the growth of the surface droplets follows the constant contact angle mode,25 the height of the droplets, Hd, varies with the droplet size, as illustrated in Figure 2b. The static contact angle of octanol (θ) on the substrate in water is 42 ± 3°. Thus, with known lateral diameter (L) of the droplets from the morphological measurements, the height of the droplets can be calculated by18
and measurement correction in the Supporting Information). Here, the oil−water partition coefficient of a compound is defined by7 Ko‐w =
Co Cw
(1)
where C0 is the concentration of the compound in oil phase and Cw is the one in water phase at the partition equilibrium. Figure 2a shows the sketch of the path of the detecting light through the channel. The upright microspectrophotometer
Hd =
(5)
We detected the absorbance of R6G in the droplets of different sizes and the R6G remaining in the water phase. Figure 2c reveals the strongest absorption of R6G at 500−550 nm. As expected, the weakest peak intensity obtained in the water phase suggests the low concentration of R6G in aqueous solution. Because more R6G partitioned into the droplets, the fraction εoCoHd in eq 3 is dominant. As such, the increase of droplet size leads to the increase of the absorbance Ad. We calculated the log Ko‑w from each size of the droplets by eqs 5 and 4, as shown in Figure 2d. The similar value of log Ko‑w regardless of the droplet size suggests the partitioned concentration of R6G in the droplets is consistent. The averaged log Ko‑w of R6G from these measurements is 2.71 ± 0.1, which is similar to the reported quantity in the literature, 2.69 ± 0.18. This result verifies our calculations in eqs 1 and 5. In the following section, we chose the droplets of 30−50 μm in diameter for the detection of log Ko‑w. The result of log Ko‑w is dependent on the linear range of the quantification equipment. Figure 2e shows the plot of the log Ko‑w as a function of the concentrations of R6G in its original solution to reveal the range of applicable concentration of R6G. The value given in the literature is labeled by the gray shaded area. Thus, the suitable concentration range of R6G for this method is confirmed to be from 1 × 10−3 to 0.05 mM. To demonstrate the reliability of this method, we detected log Ko‑w for several other pollutants and drug molecules. Figure 3a shows the representative spectra obtained from the droplets
Figure 2. (a) Sketch of the light path on the substrate. (b) Sketch of the constant contact angle mode of the surface droplets with different diameters. (c) The absorbance spectra of the R6G in the droplets of different sizes and in the water phase. (d) The log Ko‑w calculated from the droplets of different sizes. The original concentration of R6G in aqueous solution is 0.01 mM. (e) The log Ko‑w obtained from different R6G original concentrations in the aqueous solution.
acquired the absorbance signals by the reflected light from two locations: the droplet (Ad) and the liquid solution above the bare substrate (Aw). According to Lambert−Beer’s law,24 Aw is determined by
A w = εw CwH w
1 − cos θ L 2 sin θ
(2)
where εw is the molar absorption coefficient of R6G in water and Hw is the path length of the water phase, which is predetermined by the height of the flow channel. The absorbance acquired from the droplet location (Ad) consists of a combined signal of both the octanol droplet and the water phase above the droplet. Therefore: Ad = εoCoHd + εw Cw(H w − Hd)
Figure 3. (a) The UV−vis spectra obtained on the surface droplets compared with the one from the water phase for three different substances. (b) The different UV−vis spectra of sulfamethoxazole effected by the addition of salts.
(3)
where εo is the molar absorption coefficient of R6G in octanol and Hd is the height of the droplet and also the effective path length of the octanol phase. Therefore, the partition coefficient can be determined by
yz C ε ij A − A w H w Ko‐w = o = w jjj d · + 1zzz j z Cw Hd εo k A w (4) { We predetermined εw and εo of R6G, which are 1.12 and 1.04 × 105 L·mol−1·cm−1, respectively. Therefore, in the following studies, we calculate the log Ko‑w with the assumption of εw = εo in eq 4. If required, ε can be measured by the microspectrophotometer simply through introducing the
and water phase for three substances of different hydrophobicity: triclosan, carbamazepine, and caffeine. In addition, we also studied the partition behaviors of drug molecules influenced by neutral salts and buffers, as shown in Figure 3b as an example. Table 1 summarizes the log Ko‑w obtained from the droplet-based method compared to the reported values of the corresponding analytes. Most of the values are in high accordance with the literature values. The addition of salts and buffer will not influence the accuracy of this method, C
DOI: 10.1021/acs.analchem.9b02586 Anal. Chem. XXXX, XXX, XXX−XXX
Letter
Analytical Chemistry Table 1. Log Ko‑w of the Analytes Based on the Surface Droplet Method Compared to Values in the Literature name triclosan rhodamine B carbamazepine
concentrationa (mM) 0.2 ∼1 × 10−3 to 0.08 0.01−0.1
sulfamethoxazole
0.01−0.5
gallic acid caffeine
0.01−0.5 0.01−0.5
log Ko‑w, this work
log Ko‑w, literature
>3.5b 2.35 ± 0.2
4.7626c 2.2827
1.79 ± 0.18 1.58 ± 0.2 (0.1 M NaCl) 1.73 ± 0.21 (0.1 M citric buffer) 1.71 ± 0.15 (0.1 M PBS) 0.55 ± 0.15 0 (0.1 M NaCl) 0 0
1.5128 1.61d 1.53d 1.63d 0.8929 0.21d 0.1230 −0.0731
Figure 4. (a) Fluorescence images showing the partition behaviors of FR with change in solution pH and the corresponding threshold image in the bottom row. Scale bar: 20 μm. (b) The absorbance spectra of FR in droplets and the water phase with three different levels of pH. (c) Plots of the measured log Do‑w of FR as a function of pH compared to the reported value from Kasnavia et al.,32 Oba et al.,33 and Grimes et al.34 (d) Plots of the measured log Do‑w of MB as a function of pH compared to the reported value from Da Silva et al.35 The original concentration of FR and MB were both 0.01 mM in the aqueous solution.
a
The detected concentration of the analytes in original aqueous solution by the droplet method. bConcentration in the water phase is too low to be detected; value estimated by the obtained detection limit of triclosan in the water phase. cEstimated value by fragment calculation. dValues obtained by shake-flask-based UV detection in this study.
suggesting its further application in the study of ionic effect on drug molecules. We note that for the substance which partitioned almost equally in both the octanol and water phase, e.g., caffeine, the absorbance spectra obtained from the droplet and from the water phase are similar, yielding log Ko‑w = 0. As the path length in oil phase is much smaller than that in water phase, the slight difference of the concentration between the two phases can be hardly discriminated by the spectrophotometer. However, as most of the important pharmaceutical compounds and pollutants are highly hydrophobic with log Ko‑w larger than 1.5, this slight deviation of the results does not provide significant influence in defining the hydrophobicity of the compounds. For the ionized compounds which can be dissociated in the partitioning process, the partition coefficient is referred to as a distribution coefficient, log Do‑w, and is measured as a function of pH.4 In this work, our approach was also applied to the fast detection of the effect of pH on the distribution coefficient of ionized molecules with a demonstration using fluorescein (FR). The fluorescence images in Figure 4a show the distribution behaviors of FR in three different pH solution systems along with the corresponding threshold images revealing the intensity contrast of the droplet and water solution. With the increase in pH, the decrease of droplet intensity indicates the distribution of ionized FR shifted from octanol phase to water phase. The spectra shown in Figure 4b align with the behavior of FR observed in Figure 4a. The log Do‑w was calculated and plotted against the pH of the solution, as shown in Figure 4c. The results suggest good agreement with the reported values from literature.32−34 This method further performed the detection of log Do‑w to another dye/ pollutant molecule, methylene blue (MB). As shown in Figure 4d, the values of log Do‑w against the pH from 3 to 10 for MB obtained with the droplet method were compared with the ones from the shake-flask method and the electrochemical method in the literature.35 The change of log Do‑w shows the same trend for these three methods, indicating that the MB
distributes from the water to octanol phase with the increase of pH. On the basis of the measurements of partition and distribution coefficients, we now confirm the accuracy of the droplet-based method. This method can be used for the measurement of log Ko‑w ranging from −1.5 to 3.5. When log Ko‑w ≤ −1 or ≥0.7, the results of the droplet method are in high accordance with the shake-flask method. Whereas in the range of −1< log Ko‑w < 0.7, the droplet method can only acquire a log Ko‑w of 0. This limitation of our method could be solved by reducing the height of the water layer after droplet formation, increasing the contact angle of the droplet by surface modification, and improving the resolution of the equipment in the future. To summarize, we have shown a surface droplet-based platform for rapid and reliable determination of oil−water partition/distribution coefficient. A continuous solvent displacement in a flow chamber allows for an automated process for the formation of surface femtoliter droplets and online screening of a series of compounds with different hydrophobicity. When integrated with a microspectrophotometer, the concentrations of the target analytes in oil droplets and in the water phase are determined in situ. We demonstrated the reliability of this method by determining the octanol−water partition coefficient for 7 types of dye and drug molecules and the distribution coefficient for 2 chemicals with the solution pH from 3 to 10. The data gathered on our droplet platform show close agreement with literature values and present low error variation. Compared to the traditional shake-flask method, this method reduces the consumption of solvent to microliters (∼50 μL) and requires only ∼3 min for completing the detection. This solvent exchange-based approach is versatile, which can be further used in the screening of partition behaviors of drug molecules effected by temperature and in many other types of oil−water systems. The simple, green, time-effective, automated, and accurate features of this D
DOI: 10.1021/acs.analchem.9b02586 Anal. Chem. XXXX, XXX, XXX−XXX
Letter
Analytical Chemistry
(15) Herrera-Herrera, A. V.; Asensio-Ramos, M.; Hernández-Borges, J.; Rodríguez-Delgado, M. Á . TrAC, Trends Anal. Chem. 2010, 29 (7), 728−751. (16) Mashaghi, S.; Abbaspourrad, A.; Weitz, D. A.; van Oijen, A. M. TrAC, Trends Anal. Chem. 2016, 82, 118−125. (17) Shang, L.; Cheng, Y.; Zhao, Y. Chem. Rev. 2017, 117 (12), 7964−8040. (18) Lohse, D.; Zhang, X. Rev. Mod. Phys. 2015, 87 (3), 981−1035. (19) Zhang, X.; Lu, Z.; Tan, H.; Bao, L.; He, Y.; Sun, C.; Lohse, D. Proc. Natl. Acad. Sci. U. S. A. 2015, 112 (30), 9253−9257. (20) Lu, Z.; Schaarsberg, M. H. K.; Zhu, X.; Yeo, L. Y.; Lohse, D.; Zhang, X. Proc. Natl. Acad. Sci. U. S. A. 2017, 114 (39), 10332− 10337. (21) Yu, H.; Peng, S.; Lei, L.; Zhang, J.; Greaves, T. L.; Zhang, X. ACS Appl. Mater. Interfaces 2016, 8 (34), 22679−22687. (22) Li, M.; Dyett, B.; Yu, H.; Bansal, V.; Zhang, X. Small 2019, 15, 1804683. (23) Lahnstein, K.; Schmehl, T.; Rüsch, U.; Rieger, M.; Seeger, W.; Gessler, T. Int. J. Pharm. 2008, 351 (1−2), 158−164. (24) Kahl, G. Dict. Genomics, Transcr. Proteomics 2015, 1−1. DOI: 10.1002/9783527678679 (25) Bao, L.; Werbiuk, Z.; Lohse, D.; Zhang, X. J. Phys. Chem. Lett. 2016, 7 (6), 1055−1059. (26) McAvoy, D. C.; Schatowitz, B.; Jacob, M.; Hauk, A.; Eckhoff, W. S. Environ. Toxicol. Chem. 2002, 21 (7), 1323−1329. (27) Toropainen, E.; Ranta, V.-P.; Talvitie, A.; Suhonen, P.; Urtti, A. Invest. Ophthalmol. Vis. Sci. 2001, 42 (12), 2942−2948. (28) Scheytt, T.; Mersmann, P.; Lindstädt, R.; Heberer, T. Water, Air, Soil Pollut. 2005, 165 (1), 3−11. (29) Nguyen Dang Giang, C.; Sebesvari, Z.; Renaud, F.; Rosendahl, I.; Hoang Minh, Q.; Amelung, W. PLoS One 2015, 10 (7), 1−24. (30) Yalkowsky, S. H.; Valvani, S. C.; Roseman, T. J. J. Pharm. Sci. 1983, 72 (8), 866−870. (31) Lombardo, F.; Shalaeva, M. Y.; Tupper, K. A.; Gao, F. J. Med. Chem. 2001, 44 (15), 2490−2497. (32) Kasnavia, T.; Vu, D.; Sabatini, D. A. Groundwater 1999, 37, 376−381. (33) Oba, Y.; Poulson, S. R. Geochem. J. 2012, 46 (6), 517−520. (34) Grimes, P. A.; Stone, R. A.; Laties, A. M.; Li, W. Arch. Ophthalmol. 1982, 100 (4), 635−639. (35) Da Silva, J. S.; Junqueira, H. C.; Ferreira, T. L. Electrochim. Acta 2014, 144, 154−160.
unique surface droplet-based platform would be of great significance in advancing the development of future highthroughput devices.
■
ASSOCIATED CONTENT
* Supporting Information S
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.9b02586.
■
Experimental details, information and method correction of the microspectrophotometer; absorbance signal influenced by the size of the droplets; and the droplets influenced by ethanol contents and addition of salts (PDF)
AUTHOR INFORMATION
Corresponding Authors
*E-mail:
[email protected]. *E-mail:
[email protected]. ORCID
Miaosi Li: 0000-0003-4802-3714 Brendan Dyett: 0000-0002-8417-2736 Xuehua Zhang: 0000-0001-6093-5324 Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors acknowledge the support from Vice-Chancellor’s Postdoctoral Fellowship from RMIT and the Natural Science and Engineering Research Council of Canada (NSERC). This work was performed in part as the RMIT Micro Nano Research Facility (MNRF) in the Victoria node of the Australian National Fabrication Facility (ANFF) and Canada Research Chairs Program. We also acknowledge Mr. Gilmar Arends for the proofreading of this manuscript.
■
REFERENCES
(1) Leo, A.; Hansch, C.; Elkins, D. Chem. Rev. 1971, 71 (6), 525− 616. (2) Klopman, G.; Zhu, H. Mini-Rev. Med. Chem. 2005, 5, 127−133. (3) Krop, H. B.; van Velzen, M. J. M.; Parsons, J. R.; Govers, H. A. J. Chemosphere 1997, 34 (1), 107−119. (4) Krämer, S. D. Pharmacokinetic Optimization in Drug Research. Verlag Helvetica Chimica Acta: Zurich, 2001, 401−428. (5) Andersson, J. T.; Schräder, W. Anal. Chem. 1999, 71 (16), 3610−3614. (6) Baka, E.; Comer, J. E. A.; Takács-Novák, K. J. Pharm. Biomed. Anal. 2008, 46 (2), 335−341. (7) Danielsson, L.-G.; Zhang, Y.-H. TrAC, Trends Anal. Chem. 1996, 15 (4), 188−196. (8) Gluck, S. J. Anal. Chim. Acta 1988, 214, 315−327. (9) Gerhardt, G.; Adams, R. N. Anal. Chem. 1982, 54 (14), 2618− 2620. (10) Finizio, A.; Vighi, M.; Sandroni, D. Chemosphere 1997, 34 (1), 131−161. (11) Bodor, N.; Gabanyi, Z.; Wong, C. K. J. Am. Chem. Soc. 1989, 111 (11), 3783−3786. (12) Meylan, W. M.; Howard, P. H. J. Pharm. Sci. 1995, 84 (1), 83− 92. (13) Marine, N. A.; Klein, S. A.; Posner, J. D. Anal. Chem. 2009, 81 (4), 1471−1476. (14) Poulsen, C. E.; Wootton, R. C. R.; Wolff, A.; Demello, A. J.; Elvira, K. S. Anal. Chem. 2015, 87 (12), 6265−6270. E
DOI: 10.1021/acs.analchem.9b02586 Anal. Chem. XXXX, XXX, XXX−XXX
