Protein landscapes for medical genetics
The gene from the jellyfish made it possible to clarify the mechanisms of protein evolution
Biologists from Russia, the USA, the Czech Republic, Israel and Spain for the first time were able to measure the interaction of many mutations in a single protein molecule – for this they studied tens of thousands of mutants of green fluorescent protein from the jellyfish Aequorea victoria. The results, published in the journal Nature (Sarkisyan et al., Local fitness landscape of the green fluorescent protein), clarify the mechanisms of protein evolution and allow a deeper understanding of why the effects of mutations depend on the genetic context in which they occurred.
Interaction between different events can lead to consequences that greatly exceed the effect of each event. For example, a slightly leaking gas tank in your car or a weak current on its body separately are not critical breakdowns: they will completely allow you to get to the nearest workshop to repair the car. If these breakdowns occur simultaneously in the same car, then their joint action can be fatal.
"We were interested in the question of how mutations accumulating in a protein interact with each other, and how often this can occur in protein evolution," says Karen Sargsyan, an employee of the Institute of Bioorganic Chemistry of the Russian Academy of Sciences (IBH), the first author of the article. – We came up with a way to simultaneously measure the functionality of tens of thousands of mutants of one protein and used it to determine how the effect of mutations on the brightness of a green fluorescent protein depends on the presence of other mutations in the gene.
Sarkisyan and his colleagues from MIPT, Moscow State University, Nizhny Novgorod Medical Academy, the Institute of Protein and other scientific organizations in Russia, Spain, the USA, the Czech Republic and Israel studied the so-called "fitness landscape". Biologists use this metaphor to represent the evolution of organisms as a walk through a landscape in which each point of space corresponds to a certain genotype, and its height is determined by the fitness of this genotype.
Until recently, scientific methods did not allow obtaining enough experimental data to judge the structure of such landscapes. In the new work, thanks to an original experimental approach, researchers were able to look at the fitness landscape of a whole protein for the first time.
In order to measure the functionality of mutants, scientists forced mutant genes to work in the bacterium Escherichia coli, and then used an automatic device – a cell sorter – to sort cells into eight tubes, depending on the brightness of their fluorescence. Reading the DNA of mutant genes from each test tube and subsequent data analysis made it possible to compare the fluorescence brightness of each mutant with its genotype.
– In fact, this is the first experimental illustration of the concept of the fitness landscape, which was invented 85 years ago, – says Karen. "After measuring the brightness of fifty thousand mutants, we were finally able to look at what this landscape actually looks like for a particular protein.
a. Illustration of the concept of the fitness landscape of a green fluorescent protein. The green dot in the center is an unmutated protein, the dots on the circles are mutant variants with one, two and three different amino acids. The color reflects the phenotype (individual properties) of the mutant: green – the mutant glows brightly, gray – the mutant does not glow. The arrows reflect possible routes of movement across the fitness landscape.
b. Visualization of all the data obtained in the work in one picture.
The sequence of the green fluorescent protein is depicted as a circle: each small sector represents one amino acid position. The farther the circle is from the center, the more mutations the protein contains. The proportion of green in each sector reflects the proportion of functional mutants.
Scientists have found that only every fourth amino acid substitution is neutral, while most mutations negatively affect the functionality of the protein. At the same time, as in the example with the car, if one mildly harmful mutation is already present in the gene, the negative effect of subsequent mutations will be aggravated, leading to a complete loss of functionality by the protein much earlier than when the effects of mutations are independent of each other.
The researchers also managed to clarify the probable biophysical mechanisms of mutation interaction.
– We expected to see that amino acids located close to each other in the protein structure would influence each other. Instead, it turned out that pairs of interacting mutations are distributed along the protein structure, at first glance, quite randomly. We modeled the effect of mutations on the folding energy and found out that, probably, when the total load of all accumulated mutations exceeds a certain threshold – in the range of 7-9 kcal/mol – the protein simply stops folding, and the fluorescence disappears. Such a mechanism of "accounting" for mutations can help evolution to effectively weed out gene variants with mildly harmful mutations, – concludes Dmitry Bolotin, an employee of IBH, one of the key authors of the article.
The strong interaction effect of mutations that the authors found in the fluorescent protein may be important for related fields of science. One of the main tasks of modern medical genetics is the study of diseases with a complex genetic component, for example, various diseases of the cardiovascular system.
Twenty years ago, many expected that by reading the DNA sequence, it would be possible to find several specific gene changes that would explain the bulk of such diseases. At the moment, it is obvious that such diseases are determined by a large number of changes in a large number of genes. The question of the interaction of these changes and the impact on human health remains open. The new work on a fluorescent protein gives geneticists grounds to start searching for the effect of mutation interactions in complex polygenic diseases.
Experimental study of the fitness landscapes of various proteins may have another practical result. Using artificial neural networks trained only on a part of the data obtained, the researchers were able to accurately predict the brightness of the fluorescence of mutant proteins, which the neural network did not encounter during the training process.
– If we assume that the landscapes of other proteins have a similar structure, then in the future the knowledge gained from the example of this protein will allow us to evaluate the functionality of mutant forms of other proteins using only computer prediction, – adds Dinara Usmanova, a graduate of MIPT, co-author of the work.
The application of machine learning to data on fitness landscapes may prove to be an effective means of finding successful combinations of mutations and will accelerate the development of proteins that are practically important for biotechnology and medicine.
Portal "Eternal youth" http://vechnayamolodost.ru
13.05.2016