WO2006013109A1

WO2006013109A1 - Method for testing a substance interacting with a target molecule

Info

Publication number: WO2006013109A1
Application number: PCT/EP2005/008510
Authority: WO
Inventors: Hanna Jankevics; Michael Prummer; Horst Vogel
Original assignee: Epfl Swiss Federal Institute Of Technology-Lausanne
Priority date: 2004-08-06
Filing date: 2005-08-05
Publication date: 2006-02-09

Abstract

The present invention relates to a method for testing a substance interacting with a target molecule in a cell by fluorescence correlation spectroscopy (FCS) by adding a test substance to the cell and determining a distribution of diffusion coefficients for the fluorescently labelled target molecule. Preferably, the target molecule is a nuclear hormone receptor, e.g. the estrogen receptor.

Description

Method for testing a substance interacting with a target molecule

Description

The present invention relates to a method for testing a substance interacting with a target molecule in a cell by fluorescence correlation spectroscopy (FCS). Preferably, the target molecule is a nuclear hormone receptor, e.g. the estrogen receptor.

US 2002/0072076 describes a determination of the interaction of a substance with a hormone receptor in a cell by FCS wherein alterations of the diffusion coefficient of a fluorescently labelled hormone receptor in the absence or in the presence of a test compound is determined. The method allows determination whether or not the test substance is a substance interacting with the hormone receptor. The method, however, does not allow a clarification of the effect of the test substance on the hormone receptor.

In order to overcome the deficiencies of the prior art, the present invention provides a method for testing a substance interacting with a target molecule in a cell, comprising: a) providing a cell containing a target molecule carrying a fluorescent label, b) adding a test substance to the cell and determining a pattern or distribution of diffusion coefficients for the fluorescently labelled target molecule by FCS, c) optionally comparing the pattern of diffusion coefficients obtained in step

(b) with a reference pattern/distribution and d) optionally classifying the effect of the test substance on the target molecule based on the pattern of diffusion coefficients obtained in step

(b).

The target molecule may be any molecule, which can carry a fluorescent label, e.g. a peptide, polypeptide or nucleic acid, e.g. DNA or RNA.

Preferably, the target molecule is a polypeptide or protein, e.g. a hormone receptor such as the estrogen receptor (ER). The target molecule may be located in the cytoplasm, in the nucleus, in an organelle or on and/or adjacent to a cell membrane. The cell is preferably a living cell, e.g. a prokaryotic, or eukaryotic, e.g. an animal or plant cell. More preferably, the cell is a mammalian, e.g. human cell.

The target molecule carries a fluorescent label. Fluorescent labels can be added externally and subsequently bound to the target molecule, e.g. chemical compounds like ligands which specifically interact with biological macromolecules or macromolecular complexes or macromolecular structures on the cell surface, inside the cell membrane, or inside of the cell.

Alternatively cellular components might be labeled either chemically, biochemically, by expressing suitable mutant proteins, e.g. fusion proteins between the protein of interest and different fluorescent proteins such as GFP and YFP or, fusion proteins between the protein of interest and human alkylguanin-DNA-alkyltransferase which subsequently are selectively modified with a fluorescent label (Keppler et al. 2003, Nature Biotechnol 21 , 86-89), by introducing non-natural fluorescent amino acids by suppressor t- RNA technology, by using biotinylated proteins which can be labeled with fluorescent avidin/streptavidin, by introducing fluorescent bases into DNA/RNA molecules, by using fluorescent complementary DNA segments to label DNA and/or by introducing fluorescent antibodies which bind to the protein, DNA, RNA, or organelles of interest.

Preferred examples of labelled target molecules are fusion proteins of a fluorescent protein, e.g. GFP or YFP with nuclear hormone receptor. Specific examples of suitable labelled target molecules are:

Yellow fluorescent protein-labeled estrogen receptor, [YFP-ER],

YFP-labeled full length steroid receptor coactivator-3 (SRC-3), [YFP-FL], YFP-labeled receptor interaction domain (RID) of SRC-3, [YFP-RID], YFP-labeled point mutated YFP-ER (not interacting with SRCs), [YFP-ER L539A]

In the experiments of the present invention, the effect of unlabelled steroid receptor coactivator-1 (SRC-1) and agonist on the mobility of YFP-ER and YFP-ERL539A was tested. Further, the effect of unlabelled ligands on YFP- ER, YFP-FL and YFP-RID was tested.

According to the invention, further test systems may comprise the testing of specific nuclear hormone receptor interaction partners such as DNA, co- activator proteins, co-repressor proteins, chaperone proteins, proteasomes and proteins involved in the general transcription machinery.

The present invention is based on the finding that the determination of diffusion coefficients for the fluorescently labelled target molecule by FCS in the presence of a test substance under defined test conditions in a heterogeneous environment gives a pattern or distribution of diffusion coefficients which is characteristic for the effect of the test substance on the target molecule. The diffusion coefficient is preferably determined by comparing a measured intensity autocorrelation function with a model function assuming one or more, e.g. two diffusive components. The determination of a pattern of diffusion coefficients requires a plurality of measurements. The measurements may be carried out under identical and/or different test conditions. In addition, a number of distributions may be determined under different test conditions, e.g. by varying the concentration of the test substance, physical parameters such as temperature, pressure, irradiation (e.g. UV, VIS₁ IR), stress (e.g. chemical or mechanical stress) or environmental conditions (e.g. medium composition, presence of chemical compounds or surfaces with particular composition and/or structure) or any combination thereof, and a part of the whole set of distributions of the target molecule may be used to assess the effect of a test substance.

The distribution of the diffusion coefficients for a test compound (a compound for which the effect on the target molecule is unknown) may be compared with one or several reference patterns or distributions of diffusion coefficients. These reference patterns may be obtained by determining a pattern of diffusion coefficients with a substance having an already known effect on the target molecule.

According to the present invention, it may be determined, if a test substance binds or interacts with the target molecule at all. Additionally, the present invention allows a classification of the effect of the test substance on the target molecule, e.g. by determining the class of ligands the test compounds belong to. For example, the test compound may be. a partial or complete agonist or a partial or complete antagonist of the target molecule. Thus, the method of the present invention is particularly useful for clinical diagnostics and high-throughput drug screening.

Further, the present invention shall be explained in more detail by the following examples:

Examples

1. Methods

1.1 Sample preparations

Cell culture and transfection The different cell lines used in this study were Endothelial breast cancer cells (MCF-7), HeLa cancer cells, Human Embryo Kidney cells 293 (HEK293), and Chinese Hamster Ovary cells (CHO). HEK293, HeLa, CHO and MCF-7 cells were regularly maintained in DMEM/F-12 with Glutamax-I (Invitrogen Corporation, UK) with 2.2 % Fetal Calf Serum in 5 % CO₂ incubator at 37C. Three days before measurements the medium was changed for D-MEM/F-12 without phenol red (Invitrogen Corporation, UK) with 2.2 % charcoal treated Fetal Calf Serum. Two days before measurements the cells were seeded in Lab-Tek chambered cover glass (Nalge Nunc International Corp, Naperville, IL, USA) at density of 10,000 cells/ml. The vectors were transiently expressed using Effectene transfection kit (QIAGEN GmbH, Germany).

1.2 FCS experiments

FCS setup

FCS measurements were performed using the Confocor2-LSM510 microscope (Zeiss, Jena, Germany) equipped with an C-Apochromat 40x 1.2 water immersion objective.

All measurements were performed at room temperature. The YFP fluorescence was excited with a 488-nm line from an argon ion laser and detected through a pinhole with a diameter of 70 μm, and a 530-600 nm band pass filter. A laser power of below 3 kW/cm²for the FCS experiments. The LSM510 was used to localize cells with a low expression level, and to position the laser beam for the FCS measurement in different areas of the cell. The coordinates of the laser beam in the LSM510 images for different zooms and scan speeds were determined by recording the coordinates of the bleached spot seen on a cover glass slide containing fluorescent glue, after a short pulse of the FCS laser at maximum intensity settings. The cover slide was a kind gift from Zeiss (Zeiss, Jena, Germany).

FCS measurements

Autocorrelation curves were derived from fluctuations of fluorescence intensity measured over 5-seconds intervals.

For the intracellular experiments, normally 19 measurements were taken per sample spot and nucleus. The first measurement was skipped due to possible bleaching effects. In the intranuclear FCS measurements, care was taken in choosing the measurement position to avoid the presence of large speckles and nucleosomes in the laser beam. Measurements that were directly on speckles showed a high degree of initial bleaching, indicating that the speckles are relatively immobile components. Hence, all diffusion complexes seen in this study, are those that are located in between the larger speckles, but they may still be interacting with the speckles or exchanging with partners in the speckles.

To evaluate the variance between measurements at different locations inside the nucleus, four measurements were performed at each location. The first measurement was skipped due to possible bleaching effects and an average autocorrelation curve of the three following measurements was evaluated for the variations to improve the statistics.

In a typical FCS experiment there were approximately 10,000 to 100,000 receptors inside the nucleus. The numbers were estimated based on the number of molecules retrieved in the confocal volume in the initial FCS experiment, compared to the initial fluorescence intensity signal keeping in mind the counts per molecule (CPM) of the fluorophores.

Calibration of FCS setup

To calibrate the FCS set-up and to enable the comparison of data measured at different days, at the start of each series of measurements a calibration measurement was done using Rhodamine 6G (RhoδG) in PBS solution at pH 7.4. The RhoδG diffusion was obtained as an average of 20 consecutive measurements, at the experimental conditions used for the subsequent study. The confocal volume changed slightly for every measurement day, and in turn also the characteristic diffusion times. Therefore it was necessary to normalize the diffusion times retrieved in experiments made at different days to one standard rhodamine diffusion time. All diffusion times given in this work are scaled to a RhoδG diffusion time of 21 μs.

A value of 280 μnfVs was used for the diffusion coefficient of RhoδG in PBS solution in all further calculations. Hence, to calculate a particular diffusion coefficient the corresponding diffusion time (in seconds) replaces τ_D D=(21μs*280 μnτ7s )/τ_D [μm²/s]

The same calibration method was used for in vitro and in vivo experiments assuming that the cellular structures will not change the confocal volume dramatically. The assumption was tested by measuring the diffusion time of Rhodamine 6G in PBS above cells growing on the cover slip; the same diffusion time was measured as for RhoθG in PBS buffer in the presence and absence of cells, indicating that the assumption is valid. This is in agreement with a literature report where no changes in diffusion time and spot size were detected in measurements done on the basal or apical side of cells, indicating that t:he passage of the light through the cell does not change the confocal volume (Licht et al. 2003, Biochemistry 42, 2916-25).

FCS data analysis In the data analysis the raw data are described by a suitable correlation function. To fit the correlation function to the raw data, an iterative procedure was performed with the Levenberg-Marquardt algorithm to minimize chi².

Bleaching analysis of initial FCS experiment To estimate the amount of immobile molecules in the sample, the first intensity trace was fitted with a simple exponential decay. The amount of bleached molecules was calculated from the initial and the end point of the fit according to following relation: BF =((lrl_Θ)/li)*100 [%], where BF is the bleached fraction, Ii the initial intensity, and l_e the end intensity of the fit.

The mean BF and the standard deviation of the mean were calculated for the different sample conditions and used in further evaluations. Normally, at least ten cells were investigated of the same experimental conditions. The mean BF for the different ligands was plotted versus the ligand concentration to get dose-response curves. These dose-response curves were used to evaluate the potency of the different ligands to induce immobile ERs. A half-maximal value of the effective concentration (EC50) for the bleaching was calculated from these curves for each ligand in order to separate their individual potency to immobilize the ER inside the nuclei. Here, "immobile" corresponds to the timescale of FCS, i.e. to those particles which move so slowly that their fluorophores are bleached during passage across the detection volume.

Diffusion analysis of FCS experiments

The autocorrelation functions of YFP chimeras were fitted to one- , two-, or three-component models of free 3D diffusion, plus two dark-states for all in vivo measurements, to derive the translational diffusion coefficients and the number of molecules diffusing through the confocal volume.

The dark-states were introduced to cover for the laser- and pH-induced photophysical phenomena in the YFP (Schwille et al. 2000, Proc Natl Acad Sci U S A 97, 151-6). Both dark-states were kept fixed during the fitting procedure to values (12 μs and 80 μs) retrieved from calibration measurements with YFP alone. The different models used have been described in more detail in general reviews (Krichevsky and Bonnet 2002, Reports on Progress in Physics 65, 251-297; Editor Lakowicz, J.R 1991, Topics in Fluorescence spectroscopy, Plenum Press). All data were primarily fitted to a one diffusing particle model (1P*) to get an indication of mobility changes under different conditions, since the 1P* model corresponds to a population average of the different diffusing components weighted by their relative fractions contributing to the autocorrelation curve.

As a second step, all data were fitted with a two diffusing particle model (2P*). In the 2P* model one diffusion coefficient was kept to a fixed value, to increase the number of successful fits. The fixed value was chosen by first letting the two diffusion components free and determining the average diffusion coefficient of the fast and slow component. Thereafter the fast component was fixed at 4-5 different values around the average diffusion coefficient and the value finally chosen was the one that minimized the oscillations of the fit residues. The choice between 1P* or 2P* model was made by F-test as described in the next section.

Goodness of fit with F-test

In a first step, a visual inspection of the residuals from the fit to avoid oscillating residues around the baseline was used as a check for the quality of the fit, the residues are defined as res=(y~yϊ)/σi.

Second, to compare two different fits and decide which of the two different models is the most appropriate one, the reduced chi² (chi²r) was calculated according to chi²r= chi²/(n-p), where n is the number of points used, and p the number of free parameters in the fit and then used in the F-test (Bevington and Robinson 1992, Data reduction and error analysis for the physical sciences). Depending on the chosen confidence level and on chPr of the two models (chi²r1 , chi²r2), the F-distribution gives a limit for the ratio F = chi²r1/ chi²r2 if the two fits that can be assumed equally well for the data. If this ratio is higher than the given limit, the second model will be preferred over the first. The F-distribution can be found tabulated in a number of books, and it is also available from different software packages (e.g. Mathematica 4.2, Wolfram Research Inc., Champaign, IL, USA).

Creation of averaged diffusion coefficient histograms

In order to estimate the distributions from experimentally obtained histograms, and to avoid misleading peaks or shoulders arising from the choice of starting point of the histograms, the method of creating averaged histograms was chosen (Venables and Ripley 1997, Modern applied statistics with S-plus, Springer Verlag). An averaged histogram of 5 different realizations of the same data set was created. The five different realizations were made by changing the starting point of the binning by a fifth of the bin width, and then average over the 5 different realizations. The bin width was chosen to be at least two times the error in each data point.

When the data were analyzed with 2P* fits, the histograms are showing the free diffusion coefficient and the histograms are weighted histograms. Weighted histograms are mean histograms that are constructed in the following manner: the amplitude of the bin in these weighted histograms is made by adding the fraction of each diffusion coefficient that falls into that bin. The fraction of the diffusion coefficient is retrieved from the 2P* fit.

Diffusion coefficient histograms at different ligand concentrations In order to compare the distributions of diffusion coefficients caused by the different ER ligands in vivo and to compare data from different ligand concentrations, the 2P* model with two dark-states was used. The reason for keeping the same model for all ligand concentrations was that the change of model will change the diffusion coefficients retrieved, and hence to be able to directly compare the data they all have to be fitted with the same model.

The first diffusive particle was kept fixed to increase the number of successful fits. This is due to the sensitivity of the data fitting procedure. The fixed value was chosen based on a first run through the data leaving the two diffusion parameters free. This value was then kept the same for all ligand concentrations, because the choice of this value will affect the second diffusion coefficient as well.

Evaluation of ligand potency in affecting ER mobility

In order to estimate the different ligands¹ efficacy in affecting the ERs mobility inside the nucleus a single parameter was wanted as an estimator of the effect on the mobility. Here, the choice fell on the diffusion coefficient from the 1P* model since it is a population average of the ensemble of molecules contributing to the auto-correlation function (ACF) and their individual contribution is weighted by their relative fraction. From these data the median of the distribution of diffusion coefficients was taken as a robust estimator of the distribution and to estimate the error in this estimation we used the standard deviation of the mean in lack of better options. These values were plotted for each ligand concentration to achieve dose-response curves for the different ligands, and the EC50 values of the different ligands were retrieved from the fits to these dose-response curves. 2. Results

Figure 1 shows a discrimination of different concentrations of an added ligand (E2) by comparison of the measured distributions of the diffusion coefficient.

Figure 2 shows a discrimination of different added ligands by comparing the measured distributions of the diffusion coefficient of the receptor.

Antagonists (ICI), partial agonists (4OHT), and agonists (E2, BPA) exhibit different diffusion patterns. Even though the partial agonist 4OHT acts as an antagonist in MCF-7 cells its diffusive behavior differs significantly from that of the full antagonist ICI. This detailed discrimination goes far beyond a pure detection of binding events (Fig. 2A~D). A determination of the ligand binding curves of the different ligands to the receptor from the bleaching (Fig. 2E) and the mobility distributions (Fig. 2F).

Figure 3 shows a discrimination of the effect of different ligands on the estrogen receptor and 2 different coactivator segments, RID and FL.

By comparing the effect of the different ligands not only on the mobility of the estrogen receptor but also on the mobility of coactivators, another dimension of information is added to discriminate between the. ligands.

Claims

1. A method for testing a substance interacting with a target molecule in a cell, comprising: a) providing a cell containing a target molecule carrying a fluorescent label, b) adding a test substance to the cell and determining a distribution or pattern of diffusion coefficients for the fluorescently labelled target molecule by FCS assuming one or more diffusive components, c) optionally comparing the pattern of diffusion coefficients obtained in step (b) with a reference pattern and d) optionally classifying the effect of the test substance on the target molecule based on the pattern of diffusion coefficients obtained in step (b).

2. The method according to claim 1 wherein the target molecule is selected from proteins, polypeptides, nucleic acids, e.g. DNA or RNA, and other molecules which can carry a fluorescent label.

3. The method according to claim 1 or 2, wherein the target molecule is a hormone receptor.

4. The method according to claim 1 or 3, wherein the target molecule is located in the cytoplasm, in the nucleus, in an organelle, or on and/or adjacent to a membrane.

5. The method of any one of claims 1 -4, wherein the distribution or pattern of diffusion coefficients is determined for different test conditions, e.g. the concentration of the test substance, physical parameters such as temperature, pressure, irradiation (e.g. UV, VlS, IR), stress (e.g. chemical or mechanical stress) or environmental conditions (e.g. medium composition, presence of chemical compounds or surfaces with particular composition and/or structure) or any combination thereof.

6. The method of any one of claims 1-5, wherein the reference pattern has been obtained by determining a pattern/distribution of diffusion coefficients with a substance having an already known effect on the target molecule.