A fluorophore count reveals the right distribution
Method reveals the number of fluorescent molecules in a cluster
In an ideal world, fluorescent molecules used to label a biological sample always would be distinct from one another, making quantifying them a simple matter of counting. In the real world of fluorescence microscopy, however, fluorophores often congregate in clusters, or puncta, and determining how many of them are in puncta can be challenging.
Now researchers from the University of Washington in Seattle have developed and demonstrated an approach that reveals the number of single molecules present in a cluster once the relationship between single molecule and single puncta intensity distributions is known.
Knowing the fluorescence intensity distribution relationship of single molecules and particles can help in counting how many single molecules are in the particles. At top on the left is an image of single goat anti-mouse IgG antibodies (GAM) labeled with multiple Alexa Fluor 488; top right (A) shows each molecule circled automatically by the imaging software to define a region of interest. The plots below are intensity distributions of (B) single Alexa Fluor 488 carboxylic acid succinimidyl ester molecules, and (C) single synaptic vesicles tagged with anti-SV2 primary antibody and Alexa Fluor 488-labeled GAM >secondary antibody. For (B, C), the dashed line is the best-fit lognormal distribution to the data, and the dash-dot line is the best-fit normal distribution to the data. The distribution of the intensity data is a better fit by a lognormal distribution in both cases. Reprinted with permission of Biophysical Journal.
Daniel T. Chiu, professor of chemistry at the university, predicts that the technique could make fluorescence microscopy more quantitative by providing information on how many proteins, vesicles, organelles, DNAs, signaling complexes or other fluorescent units are in the puncta.
The scientists developed the method because other solutions to the problem of quantifying fluorophores had significant disadvantages. One potential approach is sequential single-molecule photobleaching, which, in theory, leads to a series of step decreases in fluorescence intensity as molecules are extinguished one by one. In practice, differences between molecules result in uneven steps, and several molecules sometimes photobleach at once, resulting in a sequence of fluorescence changes that cannot be tied to a definitive number of molecules.
Such ambiguities, along with the tedious nature of sequential photobleaching, persuaded the researchers to try something else.
Another possible method involves simple intensity measurements. However, this works only when large numbers of fluorophores with substantially the same intensity are present in each punctum. In many situations, though, there are only a few tens of fluorescent molecules in a cell — too low a number for the method to work. Thus, the researchers decided that intensity measurements would not be suitable.
The group, therefore, came up with its own technique. According to statistical theory, the fluorescence intensity of puncta and that of single molecules should be related, and the first should be predictable given the second. How the two are related affects the distribution of puncta intensities, and resolving that took the researchers years of work and a bit of luck. “We came up with this method purely by accident. Initially, we used the normal distribution,” Chiu said.
If the puncta intensities resulted from adding together single largely independent fluorophore intensities, the fluorescence intensity of the clusters would follow a normal distribution. Instead, the intensity showed a lognormal distribution, which meant that it peaked at a lower value and more sharply than a normal distribution. Thus, it seemed that the puncta intensity distribution was not simply the sum of independent fluorophores.
After thinking about these results, the researchers realized that the measurements could be accounted for if the intensity of each point in the region of interest encompassing a punctum were affected by its position in the image. In that case, the cluster intensity could be derived by scaling or multiplying single-fluorophore intensity. The result of what the researchers termed the multiplied distribution would be lognormal.
As for how this could happen, one possibility, Chiu noted, is that it arises from measurement artifacts. These could be the result of defocusing one point relative to another, of detector efficiency varying from one point as compared to another or because of a variety of other factors. Each of these issues might then be multiplied together, and the product would thereby dominate the intensity distribution.
Chiu recalled that it took years to come up with a conceptual framework to explain the observed lognormal distribution results, which, the group noted, others also had seen. He added that, if the source of this multiplied distribution is indeed many small effects occurring in the measurement process, it might be difficult to eradicate.
Whether intensities follow a random additive, a normal distribution or a multiplied distribution, lognormal distribution must be established for each case with care. In addition, Chiu noted another somewhat minor drawback to the technique. “It is computationally demanding in the fitting if there are many copies to be fitted, but with computers what they are nowadays, I think it can be handled just fine on a new desktop computer.”
In a demonstration of the technique, the researchers reacted biocytin labeled with Alexa Fluor 488 with the protein avidin, making use of one of the strongest known affinities between biological molecules and also testing the technique on a heavily studied biological interaction.
They used a microfluidic chamber and total internal reflection fluorescence microscopy on a home-built system consisting of a Roper Scientific CCD camera, a Coherent 488-nm solid-state diode-pumped laser, and dichroic mirror and filter from Chroma. The researchers used total internal reflection microscopy because it allowed them to use a lower laser power than other techniques. They needed only 88 μW of power during the capture of each 300-ms image. The work is detailed in the April issue of Biophysical Journal.
They used avidin bound to a single biocytin for single-molecule calibrating intensities, capturing 800 intensity values. They did the same for avidin bound to multiple biocytins, saturating the solution to reach the stoichiometric ratio of one avidin to four biocytins. From the fit to their data, they concluded that the multiplied distribution led to 95 percent of the avidins having four biocytins, whereas 5 percent had three — good agreement with the predicted binding ratio of 1:4. In contrast, the random additive approach led to 20 percent puncta with ratios of 1:5 or more and 20 percent with non-physical parameters, such as a negative number of biocytin per avidin.
The researchers are applying their technique to help understand synaptic function and are working to further refine the method. “We will continue to develop this technique and try different imaging modes, as we hope to make this method widely used for quantitative fluorescence microscopy,” Chiu said.
He added that he hoped the technique would someday be picked up by an image analysis company. It might be possible to have a software module that would determine the number of copies of a protein in an image with the click of a button.
- fluorescence microscopy
- Observation of samples using excitation produced fluorescence. A sample is placed within the excitation laser and the plane of observation is scanned. Emitted photons from the sample are filtered by a long pass dichroic optic and are detected and recorded for digital image reproduction.
MORE FROM PHOTONICS MEDIA