GT658 as a tool for studying single-molecule dynamics

Andrew J. Berglund

Norman Bridge Laboratory of Physics 12-33, California Institute of Technology, Pasadena, CA 91125

Recent technological advances in electro-optics have opened up the field of single-molecule optical biophysics. In a typical experiment, a laser is used to excite fluorescence in dye­-labelled biological macromolecules such as DNA, RNA or proteins. Detected fluorescence light then carries information about conformational fluctuations and chemical properties of individual molecules. In experiments where dynamics occur on timescales faster than a few milliseconds, fluorescence is typically detected by single-photon counters which generate an electronic pulse (with some "quantum efficiency") in response to a single incident pho­ton. (Even for very bright dye molecules, the optical power of the detected fluorescence is extremely low, rv 1O-15W.) Information about molecular dynamics is encoded in the pho­ton arrival times so that collecting and analyzing single-molecule fluorescence then becomes an electronic timing problem. Since many biological processes, from conformational tran­sitions to protein folding and gene transcription, rely on the interplay between molecular fluctuations across many orders of magnitude in time, measurement hardware with high time-resolution over many different times cales is crucial.

At Caltech, our goal is to develop and implement novel methods for exploiting the in­formation content of a sequence of photon arrival times. In our experiments, we use a home-built confocal microscope to focus laser light into a solution of short, double-stranded molecules of dye-labelled DNA. In order to observe single molecules, we use a solution with a low enough concentration that most of the time there are no molecules in the focus of the laser and occasionally a single molecule diffuses through the focus. The raw data in our experiment is a list of photon arrival times recorded by timing the TTL pulse output of our photon counters with the GT658. Typically, we collect fluorescence light from two photon counters, each associated with one channel on the GT658. We operate the GT658 in "packed" mode, so that data can be read directly off the card into a simple two-dimensional array consisting of photon arrival times and the channel where that photon was registered (see Fig. 1). In this format, the data is easily manipulated as an array in MATLAB, for example. The onboard clock provides our timing base, and data collection is initiated by a software trigger event. All of the control software is written in C and run on a desktop PC running Microsoft Windows 2000.

Time (s) Channel 1.000039819839 0 1.000067346122 0 1.000171748929 1 1.000237628742 1 1.000310206962 0 1.000417218816 1 1.000492098204 1

FIG. 1: Example of raw data read off the GT658 onboard memory. In the C code, individual time tags are read into the first dimension of an array and the corresponding channel where time tags were registered is recorded in the second dimension.

With the high time resolution of the GT658, we learn about molecular dynamics across many time scales from the same raw data. For example, a plot of the number of photons arriving in 1 millisecond intervals versus time shows "spikes" at times when a fluorescent molecule is at the focus of the imaging optics (see Fig. 2). The (average) length of time a molecule spends in that imaging volume is related to its diffusion coefficient, so data at the millisecond timescale is sensitive to molecular diffusion properties. Similarly, data at the microsecond timescale is sensitive to fluctuations in the overall conformation of the molecule, and data at the nanosecond timescale is sensitive to very fast intramolecular fluctuations and also to the electronic structure of the dye molecules (see [1], for example).

From the examples here, it is clear why our experiments require a time-interval analyzer with the capability to "time-stamp" electronic pulses. Since we record photon arrival times with the sub-nanosecond time resolution of the GT658, we can jump from slow (millisecond or second) to fast (nanosecond) timescales in our data analysis. Timing methods based on time-to-amplitude conversion measurements, while they can be sensitive to very short time intervals, do not allow this jumping between timescales. In statistical terms, TAC measure­ments are sensitive to second- order, exclusive correlations, or correlations between adjacent pairs of photons. With the GT658, we are sensitive to high-order, inclusive correlations, between the first and last event in the onboard memory (which holds> 106 time tags) and any in between.

For our current round of experiments, we are developing estimation procedures for infer­ring the dynamical state of a fluorescent molecule based on measurements of photon arrival times. Our estimators incorporate the additional information contained in high-order corre­lations. Starting with simple biological molecules, short strands of DNA, we hope to develop procedures which help in understanding the dynamics of more complex systems across many timescales.

Andrew Berglund is a graduate student in Hideo Mabuchi's Quantum Optics and Bio­physics group in the Physics and Control and Dynamical Systems Departments at the Cal­ifornia Institute of Technology. Before coming to Caltech, Andrew worked in the quantum information group at Los Alamos National Laboratory. He graduated from Dartmouth Col­ lege summa cum laude with High Honors in Physics in June 2000. He is the recipient of a 1999 Barry Goldwater Fellowship and 2000 National Science Foundation Graduate Research Fellowship.

---

[1] A. Berglund, A. Doherty, and H. Mabuchi, Phys. Rev. Lett. 89, 068101 (2002).