Multiplying electrons yields multiplexed results

Hank Hogan

In the past, researchers studying molecular dynamics could learn a lot about a little but had trouble seeing the big picture. For example, they could track the flow of molecules but only at about four spots on a membrane at a time. That made it difficult to fully understand the process.

Now researchers from the National University of Singapore and from the Genome Institute of Singapore have shown that a system that uses an electron-multiplying CCD camera can produce good results while allowing some 200 to 300 measurements in parallel. National University assistant chemistry professor Thorsten Wohland said that they used the camera because it allows multiplexing and because its high quantum efficiency and good signal-to-noise ratio produce a good signal.

The methods of choice for capturing molecular dynamics are fluorescence correlation and cross-correlation spectroscopy. In the former, researchers measure fluorescence intensity fluctuations that can arise from fluorophore diffusion into and out of a small excitation volume or that can be caused by chemical reactions involving the fluorophore. Those fluorescence changes can provide information about diffusion coefficients, molecular interactions, flow rates and, in biological systems, membrane dynamics.

The fluorescence signal from the focal volume of a laser can be detected by a 3 x 3-pixel area of the electron-multiplying CCD. With a pixel size of 16 x 16 μm, this area is about the size of a 50-μm-diameter pinhole used in standard fluorescence correlation spectroscopy under similar conditions. By autocorrelating the signal from the laser focus (left), one can determine flow speed, and by cross-correlating the signal from pixels around the laser focus, one can determine flow directions.

The instruments for fluorescence correlation spectroscopy typically employ avalanche photodetectors or photomultipliers as detectors. Such devices are fast, with an acquisition time of less than 0.2 ms. However, they typically have, at most, only four detectors, meaning that measurements can be made only in four spots simultaneously. Although it is possible to sequentially scan different points to cover, for example, an entire cell, that method isn’t ideal because the molecular situation can change during that span.

The big drawback to electron-multiplying CCDs is their acquisition time, Wohland noted. Fluorescence correlation spectroscopy requires measurements to be taken about 10 times faster than the slowest process, and the CCDs have acquisition times in the tens of milliseconds. That’s too slow for studies of molecular dynamics, but the researchers were able to speed that up considerably by exploiting aspects of the technology.

As detailed in the May 15 issue of Analytical Chemistry, they used a camera from Photometrics of Tucson, Ariz., which had an 8.2-mm square sensor consisting of a 512 x 512 array of 16-μm square pixels. The acquisition time for the full array was more than 30 ms, but the researchers took advantage of the camera’s capabilities to process the data faster for a smaller region of interest. By reducing the region of interest to a 20 x 20 array, they decreased the acquisition time to 4 ms, which Wohland noted was an important threshold.

“The fast acquisition rate of down to 4ms is just right to allow us to measure membrane diffusion, which is on the order of tens to hundreds of milliseconds, when using a diffraction-limited spot,” he said.

The rest of their setup consisted of a Zeiss inverted epifluorescence microscope, a helium-neon laser from Melles Griot, and a dichroic mirror and emission filter from Omega. The investigators sent the 543-nm beam from the laser through optics to a focal volume, collected the resulting fluorescence using the same optics, filtered out the emission and captured it with the back-illuminated electron-multiplying CCD sensor, which had a better than 90 percent quantum efficiency from 500 to 650 nm. The fluorescent spot covered a 3 x 3-pixel array of the CCD.

Wohland noted that having the focal volume spanned by multiple detector array elements opened the possibility of extracting additional information. For example, by autocorrelating the signal, the researchers could determine flow speed. By doing the same for pixels around that point, they could measure flow direction and movement, something not possible with conventional fluorescence correlation spectroscopy instruments because of the symmetry of the detection volume. The camera also allows for software binning and, thus, the determination of the correlation for different detection pinhole sizes, all from a single initial measurement. “This could allow the distinction of normal diffusion versus anomalous diffusion,” he said.

For their demonstration, the researchers performed fluorescence correlation spectroscopy on the fluorescent dye Atto565 and 0.01-μm fluorescent polystyrene beads in a highly viscous fluid that was 80 percent glycerol. Both fluorophores had excitation peaks near the 565-nm wavelength and emission peaks about 30 nm higher. They also performed fluorescence correlation spectroscopy on Chinese hamster ovary cells, targeting the growth factor receptors on the surface. They labeled these with a red fluorescent protein.

For flow measurements, they used a microchannel ~380 μm wide and 300 μm high. By varying the rate of a syringe pump, they created different flow velocities, which they measured using their setup.

For comparison, they used a standard fluorescence correlation spectroscopy instrument. They did see some differences in the number of particles detected and in the diffusion time for the fluorescent dye and beads. These, they explained, were due to differences between the two systems with regard to laser coupling into the optics, magnification and pinhole size. The two sets of measurements behaved the same, though, in terms of the effect of changes in concentration and other parameters.

Using software binning, they demonstrated multiplexing. Simulations showed that hundreds of detection volumes were possible, although in practice the resulting acquisition time might be too long and the resulting temporal resolution too large.

As for cell measurements, the new technique yielded diffusion times comparable to those obtained in a similar set of previously measured cells. Both measurement groups, however, exhibited large variations. Wohland said that these could be caused by differences across the membrane, about which the new system might be able to reveal more information by simultaneously measuring diffusion coefficients over the whole cell.

Having demonstrated that the technique works, the group is applying it to biological samples for diffusion and flow measurements. There are some aspects of the hardware, however, that Wohland would like to see improved. He would like to get better time resolution on larger subregions of the camera, allowing determination of correlation functions for an entire cell at once and reducing any photodamage problems. He’d also like to speed up data transfer, with the goal of perhaps eventually allowing online measurements.

When using multiple laser foci, crosstalk between the observation volumes makes it necessary to space the foci sufficiently far apart. On the left is an overlay of nine measurements. With a distance of 10 pixels, each binned 3 x 3, crosstalk does not lead to any influence on the measured diffusion time (blue dashed line represents the fast decay of correlation amplitudes with increasing distance from the laser focus). However, the amplitude is still influenced by intensity crosstalk (red solid line), which contributes to the uncorrelated background. As indicated by the colored squares, the more crosstalk present, the lower the amplitude. Corner points are least influenced, while the center point reaches only 40 percent of the expected amplitude.

Finally, he would like to see the crosstalk between various confocal volumes minimized. The need to space volumes far apart hinders the efficiency of the setup. “Most of the camera is actually unused,” Wohland said. “If crosstalk can be reduced, a much larger part of the camera could be used.”

Contact: Thorsten Wohland, National University of Singapore; e-mail: chmwt@nus.edu.sg.