DIABLA: A New Screening Method for the Discovery of Protein Targets Robbie Montgomery, Hanna Shay, Matthew McCarroll, and Luke Tolley* Department of Chemistry and Biochemistry, Southern Illinois University, Carbondale, Illinois 62901 Received May 13, 2008
Abstract: Dynamic isoelectric/anisotropy binding ligand assay (DIABLA) is a new method to identify proteins in a complex sample that bind to a molecule of interest. This is accomplished by first using capillary isoelectric focusing (cIEF) to separate the proteins in a capillary based on their isoelectric point. This separation is performed while the compound being tested is present in the separation buffer. When the proteins are focused, the entire capillary is scanned to identify regions of nonzero anisotropy, which are locations where the test compound is interacting with a focused protein band. DIABLA was demonstrated by observing the binding of fluorescein-tagged progesterone to an MCF-7 breast cancer cell lysate. The proteins were tagged with rhodamine to permit their observation and then focused in the presence of the tagged progesterone. Anisotropy measurements show that progesterone binds to six different proteins bands in the sample. Keywords: anisotropy • isoelectric focusing • drug target • progesterone
There is often an immense gap between basic and applied research, especially relating to biochemistry and medicine. Biochemistry generally studies very specific aspects of proteins or cells and medicine often seeks to understand the organism as a whole. A biochemist might understand the details of a certain protein pathway in a specific type of cell, but that information might not indicate when that pathway will be activated in a healthy or diseased organism. Medicine, on the other hand, will use a drug that is effective against a disease, even when it is not known how the drug actually works. To integrate our knowledge of the effects of a chemical on an organism with our understanding of biochemical pathways in a cell, it is necessary to know the protein targets of the chemical of interest. There are many drugs that have known effects but unknown modes of action. The lack of knowledge about the proteins involved hinders the development of safer, more effective drugs. In addition, a drug’s efficacy can be affected by its ability to bind to a receptor, as seen with imatinib (Gleevec), for instance.1 Imatinib is an anticancer drug that binds to a specific tyrosine kinase. It is an effective treatment for certain cancers, unless there is a mutation in the target protein that prevents binding. * To whom correspondence should be addressed. E-mail: ltolley@ chem.siu.edu.
4594 Journal of Proteome Research 2008, 7, 4594–4597 Published on Web 08/16/2008
Identifying unknown protein targets for specific test compounds is currently very difficult2,3 and there is no good systematic approach for doing so.4-6 If the target is known then a molecule with which it interacts can usually be found, but identifying an unknown target is much more difficult. There are various approaches that are used to study and discover interactions, but each has limitations. A recent method using self-assembled protein microarrays7 shows great promise in studying protein interactions. The proteins are synthesized in situ, eliminating many of the problems often associated with protein microarrays. This method is, however, limited to proteins with known sequences and no post-translational modifications. Fluorescence anisotropy has been widely used to study molecular interactions, particularly in biological systems.8,9 It is sensitive, fast, and can detect a wide range of binding constants. Though fluorescence anisotropy could be used to measure molecular binding in a mixture of proteins, the resulting value would simply indicate the degree of binding of the protein mixture as a whole. By itself, fluorescence anisotropy will not be able to discover the binding target of a molecule. Chromatographic affinity methods10,11 are widely used but restricted to isolating proteins that have a certain range of binding affinities. The retention of any affinity-selected protein must be high enough to separate it from the large number of other proteins, and yet be low enough that it can be eluted at a later time. This, in conjunction with the problems encountered in tethering molecules to supports, make affinity isolation impractical for any high-throughput research. Affinity methods based on capillary electrophoresis (CE) have been used to study binding in complex mixtures of proteins in a variety of ways. Equilibrium methods, such as affinity CE12-14 and fluorescence anisotropy CE (FACE), are based on having one of the binding partners present in the separation buffer while the other is separated. Nonequilibrium methods, such as affinity probe CE (APCE), mix the binding partners and then separate the bound and free components. Although equilibrium methods are able to analyze binding events that are very rapid, the nonequilibrium methods are generally limited to systems with low disassociation rates because of the required CE separation time. These methods have several advantages, such as the high resolution of capillary electrophoresis and the possibility of providing conditions similar to those encountered in vivo. In this paper, we describe a method similar to FACE, but using capillary isoelectric focusing (cIEF) rather than capillary 10.1021/pr800354f CCC: $40.75
2008 American Chemical Society
technical notes
DIABLA: Screening Method for the Discovery of Protein Targets electrophoresis to separate proteins. This combination is referred to as DIABLA (dynamic isoelectric anisotropy binding ligand assay) for convenience. Both DIABLA and FACE use fluorescence anisotropy to detect interactions between a protein and a fluorescent test molecule. In the case of DIABLA, the proteins are first separated according to their isoelectric point while in the presence of the test compound. The anisotropy of the test compound at each focused protein band is then measured. A nonzero value indicates that the protein in that band is binding to the compound. DIABLA overcomes many of the limitations with regard to the required binding strength and kinetics encountered in other methods, due to the high resolution of the cIEF separation and the sensitive fluorescence anisotropy measurements. This paper outlines the instrumental configuration and supporting data demonstrating DIABLA’s capabilities.
Experimental Methods Chemicals. Acetic acid, ammonium hydroxide, hydroxypropyl cellulose (HPC, avg. MW 100 000), HPLC grade dimethylsulfoxide (DMSO), HPLC grade water, bovine serum albumin (BSA), and phosphate buffered saline (PBS, pH 7.4) were obtained from Fisher Scientific (Fairlawn, NJ). Ampholytes (Pharmolyte 3-10 for IEF) were obtained from Amersham Biosciences (Piscataway, NJ). Tetramethylrhodamine-5-(and 6)isothiocyanate (TRITC), and progesterone receptor competitor assay (green) were obtained from Molecular Probes (Eugene, OR). Sample Preparation. Fluorescently labeled protein solution was prepared according to manufacturer’s protocol. Briefly, BSA was dissolved in water to a final concentration of 6 µg/ mL, and the progesterone receptor was dissolved in PBS to 3 µg/mL. TRITC was dissolved in DMSO at 1 mg/mL. The TRITC solution was added to the protein samples (∼1:30 ratio), and incubated in the dark for 2 h; 100 nmol of the fluorescent progesterone ligand, tagged with a fluorescein group by the manufacturer, was added to the mixture of proteins. The MCF-7 cells were cultured according to standard protocols. Briefly, cells were grown in 75 mm culture dishes in Dulbecco’s modified eagle medium containing 10% Fetal Bovine Serum and antibiotics. When they reached ∼75% confluence (about 3-5 million cells), they were dissociated from the dish with trypsin and centrifuged to a pellet. The cells were lysed using a standard RIPA buffer containing protease and phosphatase inhibitors on ice for 30 min and the soluble proteins were collected after centrifugation. The liquid was then filtered with a 0.2 µm filter and stored at -30 °C. Before use, the lysates were tagged with TRITC as previously described. Focusing Capillaries. 75 µm i.d., 360 µm o.d. fused silica capillaries (Polymicro, Phoenix, AZ) were coated with hydroxypropyl cellulose by filling the capillaries with a 5% HPC solution and then baking them at 200 °C for 45 min.15 This effectively eliminates electroosmotic flow in the capillary. The TSU line of capillary used has a transparent coating rather than polyimide, eliminating the need to remove the coating for optical detection. Instrumentation. cIEF. The high voltage power supply used was fabricated in house. It contains five 10 kV supplies (Ultravolt, Ronkonkoma, NY) and each voltage is individually controlled by a 16-bit digital-to-analog converter through a computer using software written in LabView (National Instruments, Austin, TX). The catholyte consisted of 0.2 M acetic acid and the anolyte was 0.5% ammonium hydroxide in water. A
Figure 1. Block diagram of capillary isoelectric focusing combined with fluorescence anisotropy (DIABLA) for finding the proteins with which a test compound interacts.
10 cm section of coated capillary was used. This capillary was then filled with a mixture of the sample (containing the proteins and the fluorescent progesterone) and 1% carrier ampholytes prior to analysis. Fluorescence Anisotropy. Fluorescence emission was viewed using an Olympus inverted microscope with a 4× long-range objective in an epifluorescence configuration. Excitation for the TRITC tagged proteins was at 544 nm with emission at 572 nm. Excitation for the fluorescent progesterone was at 494 nm with emission at 520 nm. A polarization filter (Hoya, Long Beach, CA) was used for anisotropy measurements and a digital camera (Cannon PowerShot G7) was attached to the microscope for recording images and video.
Results and Discussion DIABLA Instrumentation. A small fluorescent compound in free solution will generally have an anisotropy value close to zero due to its high rotational rate, as compared to its fluorescent lifetime. This means that the amount of light emitted having the same polarization as the excitation will be the same as the amount having an perpendicular polarization. When it is bound to a protein, however, its rotational rate will decrease much below the fluorescent lifetime, causing an observable change in its anisotropy. One of the advantages of anisotropy is that measurement does not depend on the binding kinetics. The increase in anisotropy will be measurable even if the binding kinetics are very fast because of the relative speed of the fluorescence lifetime, as compared to the time required for a molecule to move into and out of the binding site. This is an advantage over methods, such as APCE, that require long dissociation times. To demonstrate the concept of discovering the target of a small molecule using DIABLA, a cIEF system was integrated onto the stage of a fluorescence microscope as shown in Figure 1. The microscope was operated in epifluorescence mode, with a polarizer placed between the microscope objective and the capillary, as illustrated, to record only the parallel light emission. Since both the parallel and perpendicular emission intensities are not measured the true anisotropy value is not actually measured. The parallel light intensity will increase as the anisotropy increases, which does permit us to distinguish areas with higher anisotropy than the background. True anisotropy measurements would be possible with some additions to the instrument, such as a rotating polarizer on the Journal of Proteome Research • Vol. 7, No. 10, 2008 4595
technical notes
Montgomery et al.
Figure 2. Depiction of expected and observed anisotropy response of two proteins separated in a capillary using isoelectric focusing in the presence of fluorescently labeled progesterone. The red photographs show the focused, labeled proteins and the green only show regions where the progesterone has a high anisotropy. This method can be used to find the proteins in a mixture that bind to a specific test compound.
detector to permit both parallel and perpendicular light intensities to be measured. This type of modification will be performed at a future date. Whole column imaging16 will also be investigated. As a simple test case, human progesterone receptor (HPR) and bovine serum albumin (BSA) were analyzed using DIABLA. Both HPR and BSA were fluorescently labeled with a rhodamine tag prior to analysis and the progesterone ligand was labeled with fluorescein. The results of this experiment are presented in Figure 2. Both focused proteins are visible, but only the HPR causes the progesterone to have an anisotropy that significantly differs from the background. This clearly shows that progesterone binds strongly to HPR and not to BSA. DIABLA was then applied to detect the binding of progesterone to proteins in a cell lysate. MCF-7 breast cancer cells were used because of the known presence of HPR. The experiment was performed as before with one fluorophore for the proteins and another for the ligand. Figure 3 shows a small section of the capillary near pH 7. There are several labeled protein bands visible, but only one exhibits an increased anisotropy value. Other sections of the capillary (shown in the Supporting Information) contained a total of five additional progesterone-binding proteins, as indicated by anisotropy, with pI values ranging from ∼4-8. It is possible that this anisotropy increase could be due to nonspecific binding or pH dependent changes in the proteins, but it also could be evidence of additional progesterone receptors. One of the disadvantages of DIABLA is that the binding of each protein is measured at its isoelectric point, which is generally not at physiological pH. Studies have been performed on various proteins to determine the effect of pH on binding interactions. The results depend on the particular protein studied. Some proteins bind much better in an acidic environ4596
Journal of Proteome Research • Vol. 7, No. 10, 2008
Figure 3. DIABLA applied to MCF-7 cell lysates. The top picture shows a small section of the focused rhodamine-labeled proteins (yellow) inside of the capillary. The bottom photograph shows the anisotropy of the labeled progesterone (green), which indicates that one protein in this section binds to progesterone and that the others do not.
ment while others bind more poorly. Because of this, it is not possible to predict the effect that pH will have on binding in general. To estimate the magnitude of this problem, binding of the progesterone-HPR system was tested at multiple pH values. Using a commercial fluorescence anisotropy system, the binding was measured at six pH values ranging from 3.5 to 10. The difference between the highest measured anisotropy and the lowest was less than 50% (data not shown). Though this is a significant change that would greatly affect the calculation of a binding constant, the anisotropy is still large enough to be easily observed with DIABLA. Though this result may not be applicable to every protein, it does show that binding is not always greatly affected by the pH at which it is tested. It is likely that there will be pH-dependent changes for each protein, so the anisotropy measurements from DIABLA will not necessarily
technical notes
DIABLA: Screening Method for the Discovery of Protein Targets reflect the binding constants under physiological conditions. This is not a serious concern since the aim is not to take accurate measurements of the binding constants, but rather to determine whether or not a certain protein interacts. To prevent false positive results, all isolated fractions will undergo a second binding assay. The pH of the collected fraction will be adjusted to physiological levels and the anisotropy will be measured again. This will ensure that the observed binding is not an artifact of the pH at which it was originally measured. The pH dependence of the ligand fluorescence is something to be aware of. In this case, progesterone was tagged with fluorescein, which has rapidly decreasing yields as it becomes protonated in acidic environments. This limits its sensitivity at low pH values, but a different fluorescent tag could easily be used for finding protein targets with very low pIs. Another potential problem that occurs in isoelectric focusing is protein precipitation. Because proteins are not charged at their isoelectric point, some of them aggregate and precipitate if their concentration is high. This would interfere with their binding and cause a false negative to be observed. We have not encountered any precipitation while performing these experiments, but if it were a problem then the protein concentration could simply be reduced. These data illustrate the utility of DIABLA for discovering drug receptors, though the current setup does not easily allow for the identification of the bands of interest. This will be remedied through the integration of dynamic isoelectric focusing,17 which will enable the efficient extraction of focused bands. These protein bands will then be further analyzed and identified using additional analytical methods such as LC-MS and MALDI-MS. The identities of the collected proteins will help us to understand the way that progesterone produces all of its effects. DIABLA can be performed using different types of protein mixtures and test compounds. The test compounds have the restriction that they need to be fluorescent and have a high rotational rate compared to proteins. Many drugs have sufficient aromatic conjugation to enable native UV fluorescence, but a fluorescent tag can be added if needed to meet this requirement. To identify the interactions of the molecule of interest with the proteins, it has to be present throughout the focusing capillary. Since the pH within the capillary will change, we need to consider the effect the pH will have on the molecule itself. There are two main possibilities, the first being that the molecule is not ionized, as in the case of progesterone, in which case the electric field will not affect its position and it will remain evenly distributed within the capillary. The other possibility is that it does get ionized which will cause it to migrate in the column. In this case it will still be present in the entire capillary as long as it is present in the buffer reservoirs at both ends of the capillary. The analysis is more difficult if the molecule is neutral for part of the pH range and charged for the rest. This will create an uneven distribution of the molecule through the capillary as some of it migrates and the remainder does not. Though this is undesirable, it is not an insurmountable problem because the accumulation would not occur very quickly and
fluorescence anisotropy is fundamentally independent of analyte concentration. To avoid an uneven distribution entirely, it is also possible to perform two different analyses. One would span only the pH range where the molecule was neutral and the other would cover the range where it was charged. This approach will solve the problem. Importantly, it is entirely predictable whether this problem will occur for a given analyte. It is likely, however, that a difference in the charge state of the molecule from physiological conditions will have a large effect on its ability to bind to targets. The ability provided by DIABLA to rapidly identify interactions between a small molecule and proteins will enable researchers to understand the mode of action of drugs or other small molecules. DIABLA can also be used for other innovative experiments, such as identifying any protein receptors that are involved in unwanted drug side effects. It could also be an important tool for personalized medicine by allowing the comparison of protein binding patterns from different individuals to identify whether a certain drug could be an effective treatment for a particular person.
Acknowledgment. This research was funded by NIH/ NCI grant 1R21CA120691-01, NSF grant DBI-0754696, and a seed grant from Southern Illinois University. Supporting Information Available: Figure of DIABLA applied to MCF-7 cell lysates. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Carter, T. A.; Wodicka, L. M.; Shah, N. P.; Velasco, A. M.; Fabian, M. A.; Treiber, D. K.; Milanov, Z. V.; Atteridge, C. E.; Biggs, W. H.; Edeen, P. T.; Floyd, M.; Ford, J. M.; Grotzfeld, R. M.; Herrgard, S.; Insko, D. E.; Mehta, S. A.; Patel, H. K.; Pao, W.; Sawyers, C. L.; Varmus, H.; Zarrinkar, P. P.; Lockhart, D. J. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 11011–11016. (2) Chapal, N.; Molina, L.; Molina, F.; Laplanche, M.; Pau, B.; Petit, P. Fundam. Clin. Pharmacol. 2004, 18, 413–422. (3) Kley, N.; Ivanov, I.; Meier-Ewert, S. Pharmacogenomics 2004, 5, 395–404. (4) Lokey, R. S. Curr. Opin. Chem. Biol. 2003, 7, 91–96. (5) Ueguchi-Tanaka, M.; Ashikari, M.; Nakajima, M.; Itoh, H.; Katoh, E.; Kobayashi, M.; Chow, T. Y.; Hsing, Y. I. C.; Kitano, H.; Yamaguchi, I.; Matsuoka, M. Nature 2005, 437, 693–698. (6) Baetz, K.; McHardy, L.; Gable, K.; Tarling, T.; Reberioux, D.; Bryan, J.; Andersen, R. J.; Dunn, T.; Hieter, P.; Roberge, M. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 4525–4530. (7) Ramachandran, N.; Hainsworth, E.; Bhullar, B.; Eisenstein, S.; Rosen, B.; Lau, A. Y.; Walter, J. C.; LaBaer, J. Science 2004, 305, 86–90. (8) Xu, Y.; McCarroll, M. E. J. Phys. Chem. A 2004, 108, 6929–6932. (9) Xu, Y.; McCarroll, M. E. J. Phys. Chem. B 2005, 109, 8144–8152. (10) Nakaie, C. R.; Lanzer, D. A.; Malavolta, L.; Cilli, E. M.; Rodrigues, M. M. Anal. Biochem. 2003, 318, 39–46. (11) Zajdel, P.; Bojarski, A. J.; Bugno, R.; Jurczyk, S.; Kolaczkowski, M.; Nowak, M.; Subra, G.; Martinez, J.; Pawlowski, M. Biomed. Chromatogr. 2004, 18, 542–549. (12) McKeon, J.; Holland, L. A. Electrophoresis 2004, 25, 1243–1248. (13) Heegaard, N. H. H. Electrophoresis 2003, 24, 3879–3891. (14) Heegaard, N. H. H.; Nilsson, S.; Guzman, N. A. J. Chromatogr. B 1998, 715, 29–54. (15) Shen, Y. F.; Smith, R. D. J. Microcolumn Sep. 2000, 12, 135–141. (16) Johansson, J.; Witte, D. T.; Larsson, M.; Nilsson, S. Anal. Chem. 1996, 68, 2766–2770. (17) Montgomery, R. L.; Jia, X.; Tolley, L. T. Anal. Chem. 2006, 78, 6511–6518.
PR800354F
Journal of Proteome Research • Vol. 7, No. 10, 2008 4597
