THE IRE RNA.

Current Hall Lab Projects

 

 

 

 

 

Fluorescence Fluctuation Spectroscopy and Binding Stoichiometry of PTB. (Artem Melnykov)

 

     Our stoichiometry data describe only the size of the largest complex, a restriction of the method. The stoichiometry of the intermediate complexes is also of great interest, since these are likely to be highly populated in vivo. Fluorescence fluctuation spectroscopy (FFS) is capable of quantitatively describing the equilibrium composition of a complex, and the equilibrium concentrations of bound vs unbound components (Rigler & Elson, 2001). For these PTB experiments, the protein is fluorescently tagged for detection, and the RNA:protein complex is monitored over time to describe the fluctuations in its composition through measurement of the fluorescence intensity (brightness). Brightness analysis in fluorescence fluctuation spectroscopy is a technique complementary to fluorescence correlation spectroscopy (FCS). Rather than focusing on dynamic character of fluctuations, photon counting histogram (PCH) considers their magnitudes. Therefore, the two data analysis methods complement each other: while FCS measures diffusion coefficients and rates of chemical reactions, PCH determines brightness per molecule.

While PCH analysis is a promising technique, very few successful applications have been reported. One reason is that the resolving power of the method is not sufficient in most cases. For example, it is a challenge to resolve a mixture of monomers and dimers. The reason for that can be understood in simple terms. When fitting a histogram, we are optimizing the model parameters, brightness (q) and concentration (c), so that the model histogram approximates the measured one. It is always possible to find values for c and q such that we match the first two moments of the histogram. While higher order moments are also important for describing the shape of the distribution, they have a smaller effect and therefore such differences are harder to pick up.

One straightforward approach towards improving the resolving power of PCH is to increase the bin time used to build histograms. Traditionally histograms in brightness analysis are built using a time window that is much shorter than the characteristic decay time for the process giving rise to fluctuations. In the simplest case, this is the diffusion time as measured by FCS. This constraint on the time window has to be imposed since strictly speaking the PCH theory applies only when the molecule remains in the same excitation intensity shell during the bin time. However, intuitively a longer time window would be more appropriate for capturing properties of the sample. Such a bin time could have the same order of magnitude as the characteristic diffusion time.

 

Performance of simple PCH model. As a model system for proof of principle for the application of this new data analysis methodology, we have used DNA oligonucleotides.

For these experiments, we have used RhodamineGreen-d(T)15 and Cy3-d(T)15 as model compounds. These fluorophores differ in brightness by a factor of two, and so serve as a model system for the mixtures of monomers (1X brightness) and dimers (2X brightness) that we will need to resolve. Table 1 shows results of fitting histograms calculated with different bin times for RhG-d(T)15 (similar results were obtained for Cy3-d(T)15).  A simple one-species PCH model with the correction factor for one-photon excitation (F) was used to fit the data. As indicated by low values of χ2 for each fit, the PCH model is able to describe the shape of the curve in all cases. Essentially, this fact indicates that for any time window, we can use two parameters to describe the properties of the sample: apparent brightness per molecule and apparent number of molecules in the beam. Both these values increase with the increasing bin time as could be expected.

 

RhG-(dT)15

Tbin

C

q

F

χ2

20

1.01 ± 0.01

1.64 ± 0.01

0.45 ± 0.02

1.3

40

1.08 ± 0.01

3.05 ± 0.01

0.46 ± 0.01

1.2

100

1.33 ± 0.01

6.31 ± 0.03

0.43 ± 0.01

1.2

200

1.70 ± 0.02

10.11 ± 0.05

0.40 ± 0.01

1.3

400

2.38 ± 0.03

15.0 ± 0.1

0.35 ± 0.01

1.0

 

Table 1. The results of fitting histograms calculated from RhG-(dT)15 solution with different bin times. tdiff is 74 μs as measured by FCS.

c is the average number of molecules in the beam, q is brightness per molecule (that is, the average number of counts per molecule per time unit at given power), and T denotes the time bin used in calculating the histogram.

 

While this model works well when a single species is present, our goal is to characterize heterogeneous mixtures of PTB:RNA complexes. Two properties of these complexes will complicate interpretation of the data; one is their differences in brightness, and the other is their differences in diffusion times. Strictly speaking, PCH theory is not corrected for molecular motion, and using it to analyze histograms distorted by diffusion may result in unpredictable systematic errors. In order to eliminate this possibility, we designed a correction algorithm for PCH analysis based on the cumulant expansion of the generating function for the histogram.

 

Correction of cumulant expansion for diffusion: binning functions

It is possible to correct cumulant expansion of the PCH generating function (Qian & Elson, 1990; Muller, 2004) for diffusion and then use such correction to fit the histogram. The corrected cumulants are

 

Example graphs of the first six binning functions (Bn) are shown in Figure 2. The limit of short time where all binning functions are equal to T n corresponds to the case of stationary molecules. On the other hand, in the limit of long time, binning functions decay to 0, and any information about fluctuations is lost. It is also important to note that higher order binning functions decay faster than the Text Box:  lower order ones, and that the difference between the higher order functions diminishes.

One has to realize limitations of such an approach to generalizing the PCH model. Calculation of binning functions is a time consuming effort; therefore these functions have to be pre-calculated before fitting can be done. Also, the point spread function of the instrument influences this calculation. Finally, the effects of photophysical phenomena such as photobleaching, saturation and inter-system crossing on the apparent beam profile have not been studied in the case of PCH and may result in artifacts. In practice, a three-dimensional Gaussian beam seems to be a very good approximation for the point spread function as far as PCH is concerned, and the factor F introduced for one-photon excitation takes care of the beam non-idealities manifested through γn factors. Photophysical problems can be minimized by a judicious choice of fluorophores.

 

Performance of PCH model corrected for diffusion with different bin times. We next examined the effect of diffusion correction on the extracted parameters and compared results to those presented from the simple model. Apparently, the correction is successful in describing how the shape of the histogram depends on the binning time. It then follows that we can fit re-binned histograms globally and extract the characteristic diffusion time of the fluorophore. As shown in Figure 3, that is true and the diffusion time can be extracted together with concentration and brightness per molecule. More results of global fits of 40-100 μs histograms are shown in Table 2, for varying order of correction. Two observations are important. One is that even lowest (second) order of correction is very successful in extracting tdiff and removing bias in the values of c and q. The other observation is that the values of tdiff are not equal to those determined by FCS analysis, which are 74 μs for RhGreen-dT15 and 72 μs for Cy3-dT15.

 

Figure 3. Global fit of multiple histograms with the PCH model corrected for diffusion to the sixth order (Cy3-(dT)15). C is average number of molecules in the beam; q is brightness; Td is diffusion constant.

 

 

Table 2. The results of global fitting of histograms calculated from RhGreen-(dT)15 solutions.

RhGreen-(dT)15

Order of

correction

c

q

tdiff

F

X2

2

0.96 ± 0.01

7.03 ± 0.09

80.8 ± 1.4

0.442 ± 0.003

1.3

4

0.97 ± 0.01

6.87 ± 0.09

79.3 ± 1.3

0.430 ± 0.003

1.2

6

0.97 ± 0.01

6.97 ± 0.09

80.8 ± 1.4

0.429 ± 0.002

1.4

 

 

Characterization of a mixture of species: the effect of diffusion correction

Finally, we revisit our results for histograms collected with binary mixtures with the diffusion-corrected model. From our experience, fitting with the second-order corrected model (FIMDA/PCMH) results in unpredictable bias in the extracted model parameters. This result may seem surprising considering the success of second order correction in the case of homogeneous solutions. The failure of the FIMDA model is most likely due to the fact that all model parameters are highly correlated. Therefore, relatively small systematic errors drive the minimization algorithm towards a combination of parameters that is very far from the true set expected for the sample.

 

Table 3. Comparison of model parameters extracted by PCH models corrected to various order.

Cy3-(dT)15 : RhGreen-(dT)15

Order of

correction

c1 and c2

q1 and q2

tdiff 1 and tdiff 2

F

χ2

2

0.48 ± 0.02

0.19 ± 0.02

4.5 ± 0.4

6.9 ± 0.3

92 ± 10

62 ± 4

0.46

1.0

4

0.41 ± 0.01

0.26 ± 0.02

3.1 ± 0.2

7.3 ± 0.2

70 ± 2

73 ± 2

0.46

0.9

6

0.42 ± 0.02

0.30 ± 0.02

2.4 ± 0.2

7.2 ± 0.2

66 ± 1

72 ± 1

0.46

1.0

 

The results of global analysis of 40-100 μs histograms for the case of a binary mixture of Cy3-(dT)15: RhGreen-(dT)15 are given in Table 3. The first line of the table serves as an example of poor performance of PCMH2 model. PCMH4 and PCMH6 models are able to extract parameter values that are much closer to the expectation in this case (the last digit in the model name denotes the correction order).

In summary, PCH analysis can benefit from the theory that explicitly takes diffusion into account. With such modification, not only all major sample properties (concentration, brightness, and diffusion coefficient) can be determined in a single experiment, but also the resolving power of the method is improved. The correction for diffusion can be cast in terms of the cumulant expansion of the PCH generating function where the notion of correction order arises naturally. The corrected model allows reliable determination of sample parameters for simple solutions of fluorescent molecules and for their mixtures. It is the latter case where higher order (above two) corrections are necessary in order to obtain meaningful results.

It should be emphasized that while the approach outlined above is quite general, it requires prior time consuming calculations of the binning functions. Therefore, in the cases where some prior knowledge about the sample is available, PCMH2 can be used instead. For example, given two brightness values for the binary mixture, PCMH2 can reliably determine concentrations at which molecules are present in the mixture.

Chen,Y., Muller, J.D., So, P.T.C., Gratton, E. 1999. The photon counting histogram in fluorescence fluctuation spectroscopy. Biophys J. 77, 553-567

Muller, J.D., 2004. Cumulant analysis in fluorescence fluctuation spectroscopy. Biophys J. 86(6):3981-92.

Qian, H., Elson, E.L., 1990.  Distribution of molecular aggregation by analysis of fluctuation moments. Proc Natl Acad Sci U S A. 87, 5479-83.

Rigler, R., Elson,E.L. 2001. Eds, “Fluorescence correlation spectroscopy” Springer, Berlin.

 

 


Return to Washington University Biochemistry and Molecular Biophysics Homepage

Dr. Kathleen B. Hall (hall(at)biochem.wustl.edu)
Department of Biochemistry and Molecular Biophysics
Washington University School
of Medicine
660 South Euclid
St. Louis, MO 63110 (USA)

office: 314-362-4196
lab: 314-362-4197
or 314-747-8079
FAX: 314-362-7183
send email to kathleenhal(at)gmail.com

Last Update: December 5, 2006
This site maintained by Tom Stump
Send site related comments or concerns to tstump(at)wanda.wustl.edu