In science, it is a well known fact that in the process of applying extending theories to the real world, it is as much a question of known “how much” of something is happening, apart from what exactly is happening. Its this crucial aspect that has physicists worrying about adding never ending degrees of precision to any numbers they report (typical numbers in particle physics have up to 9 significant digits [PDF alert]). Chemists know well how important getting the right amount of catalysts is in order to make test-tube magic. Biology , unfortunately, has been the orphan child of this scientific obsession. The complexity of a single cell introduces so many variables that approaching the precision of physics is all but deemed impossible. Indeed, some have suggested that certain behaviors of living systems may really be ‘incomputable’. But perhaps this is why it is an exciting time for biology. Scientific thought in the last 20 years has morphed slowly from the search for universals to understanding the nature and consequences of variability. It is this variability that recent work (link at the bottom) from Grecco et al. at the Max Planck Institute for Molecular Physiology, published in Nature Methods seeks to unravel by giving biologists the means to do so.

 

Specifically, Grecco et al. study the flow of information in the form of phosphate groups, a staple information tag that cells attach to their proteins to switch them on , off or give them special properties. Phosphate groups are attached to proteins by enzymes called Kinases, and removed by enzymes called Phosphatases. The fine balance between the kinase and phosphatase activities determines which proteins are phosphorylated and remain so for varying lengths of time.  A lot is known about which proteins get phosphorylated under what circumstances from tedious experiments over the last years that involve extracting the protoplasm into jumbled soup and then going about detecting how much phosphate is attached to which proteins. By replacing this step with a clever optical measurement, the authors opened the door to studying these interactions in living cells. Because the method is optical, and essentially works on a microscope (albeit not a very conventional one), the authors can not only tell which proteins get phosphorylated in response to a particular stimulus , but also where exactly in the cell these proteins are located.

 

To extract this information, the authors create a library of proteins containing phosphorylatable Tyrosine residues* , each tagged with one of the now famed fluorescent proteins. They adapted a technique called ‘reverse transfection’ to create a array of tiny spots of cells (a cellular array) , each expressing a different fluorescently-tagged protein. Then they added a stimulus – Epidermal Growth Factor, or EGF in short – an extracellular signal that sets off cellular signaling eventually leading to growth of a tissue. Then they added an antibody that would specifically recognize phosphorylated Tyrosines. The antibody itself was tagged with a particular fluorescent molecule. What they ended up with is cells, glowing with a fluorescent protein they had made it express, and an antibody that attached to this protein if its tyrosine were phosphorylated.

 

Now came the optical part. How does one find out if the antibody had bound to a protein? One of the most sensitive ways of detecting binding is to use a phenomenon called Foerster/Fluorescence Resonance Energy Transfer (FRET). Fluorescent molecules absorb light of a particular color, and electrons in them become ‘excited’, storing the energy for a brief instant. Eventually though, the electron gives off this energy as light again, but because some of the energy has been lost (no process is 100% efficient) , the light it gives is of a different color. Fluorescent molecules thus have an ‘absorption spectrum’ , and an ‘emission spectrum’. What happens , though, if there is another molecule sitting right next to this excited molecule that can absorb at exactly the same color that our excited molecule is about to emit? The laws of quantum physics say that , there is a chance that resonance will occur between these molecules and the excited molecule , instead of emitting light, will give away its energy to this other fluorescent molecule, that is perfectly suited to taking up this energy. This transfer can occur however , over only a very short distance , the so called Foerster radius – such that if it does occur, in almost all cases, it means that the two molecules are less than 20 angstroms from each other. In a solution of molecules, this is only likely to happen in any significant degree if the molecules are really bound to each other. In practice, it means the following, imagine a fluorophore that absorbs blue light and gives off green light. Another absorbs green light and gives off red light. If these molecules are so close as to be bound to each other, then shining blue light, will give off some green light, but also some red light. Some of the energy of the first fluorophore (the donor )has ‘leaked’ into the second one (the acceptor).

 

Left : A schematic representation of FRET-FLIM used for detecting Tyrosine Phosphorylation. Right : The high-throughput microcopy setup for measuring FLIM in biological samples.

 

FRET has been measured usually as the brightness of the signal. If energy is leaking between fluorophores, than the acceptor is brighter than it should be, and correspondingly the donor is dimmer than it should be. This method is not very precise in finding out how many of the donor-acceptor molecules are bound, though, since there are always unbound donors that confuse the signal. The authors use a different method to quantify FRET – they measure the average amount of time the donor fluorophores stay in their “excited” state – the fluorescence lifetime. If there is an acceptor bound and energy can leak, the donor molecules should remain in their excited state for a much shorter time – in other words their fluorescence lifetime will reduce. When performed on a microscope, the method is called Fluorescence Lifetime Imaging Microscopy (FLIM). The most useful part is that the average fluorescence lifetime reduces precisely in proportion to the amount of donors bound to acceptors, and FLIM can therefore tell us how much of a protein is phosporylated and thus bound to an antibody with an acceptor fluorophore.

The authors thus combined a high-throughput screening approach, with a sensitive optical method to detect protein phosphorylation, creating Cell Array – FLIM (CA-FLIM). They now had a way to see which of many proteins were phosphorylated, to what extent, and where in the cell, at the level of individual cells, when they added a stimulant. This is a level of detail in information that has been achieved in very few instances in biology. As with all cases of a data avalanche, however, the authors quickly found that biological variability was showing its effects – not all cells responded in the same way, and conventional methods of handling the data were incapable of resolving real phenomenon from the all-pervasive effects of this variability. So they had to develop a new way of analyzing this information, grouping cells into clusters based on how much of which protein was being phosphorylated. The mathematics of this new kind of ‘global analysis’ itself forms a significant proportion of the paper.

Finally, the authors compared their results with those obtained from the study of individual proteins done with tedious conventional methods. They found their data agreed with what other scientists had painstakingly discovered over years. Of course, CA-FLIM had revealed information of those proteins and many more in a matter of hours! What CA-FLIM added to the mix in many cases is information about the spatial aspects of this phosphorylation of proteins.

One point of critique here is that in order to detect the the phosphorylation of any protein, that protein needs to be expressed in its fluorescently tagged form. While this ectopic expression is relatively easy to do, it adds an extra population of that particular protein to what the cell itself is creating from its genome, known as the endogenous protein level. Since, the endogenous protein is not fluorescent, what happens to it remains in the dark. In all studies involving ectopic expression, the implicit assumption is that the fluorescent population behaves similarly to endogenous the non-fluorescent one , and all modifications occur to the same extent on both molecules. While this is more or less an reasonable assumption for many proteins , there is no way to be absolutely sure. Indeed, the assumption is in no way ironclad and several exceptions are known. In certain model organisms, such as yeast, it is possible to get around this problem by replacing the organism’s genes with fluorescently tagged versions of those same genes, but in mammalian cells, this is not yet technically possible.

CA-FLIM now adds to the growing list of methods that reduces the time required to detect cellular protein interactions by an order of magnitude. As technical challenges in implementation and automation are overcome, and CA-FLIM is expanded to other stimuli beyond the EGF used in this study, the multi-pronged nature of this approach should provide insights, or at least shine the light on interesting cellular phenomenon – increasing our knowledge of the basic unit of life further.

 

 

*Proteins can be phosphorylated on several amino acid residues , most commonly on Serine, Threonine and Tyrosine. Tyrosine phosphorylation seems more significant in the first steps of intracellular signaling and the authors focused on this particular kind.

 

NOTE : Nachiket Vartak , the author of this post , did not contribute to the study and development of CA-FLIM, but is affiliated to the same institution where the work was done. The author declares no conflict of interest.

Editor's Selection IconGrecco, H., Roda-Navarro, P., Girod, A., Hou, J., Frahm, T., Truxius, D., Pepperkok, R., Squire, A., & Bastiaens, P. (2010). In situ analysis of tyrosine phosphorylation networks by FLIM on cell arrays Nature Methods, 7 (6), 467-472 DOI: 10.1038/nmeth.1458