First-hand news about CRISPR

"It's just very beautiful"

Alexander Ershov, N+1

A joint Russian-American group of biologists managed to discover a new type of bacterial immunity based on CRISPR. Unlike already known systems, it works only at the RNA level, which is very unusual and, potentially, may be interesting for cancer therapy. A lot of attention is now focused on everything related to CRISPR, and the flow of new articles (both scientific and in the media) is growing exponentially. We have repeatedly written about this technology, but this time we decided to focus a little more on the science behind it. And to talk not only about the principles of genome editing, but also about the latest trends in this area. All this was told to us by one of the authors of the new article, a professor at Skoltech and Rutgers University, head. laboratories at IMG and IBG RAS, Konstantin Severinov.

– Our readers have already heard about CRISPR/Cas9 technology and know something about it: that this is a new method of genome editing, that it can be used to treat hereditary (and not only) diseases, that its clinical trials are already beginning in some countries, and that this whole technology is based on the discovery of the bacterial antiviral immunity system, which works as a kind of "card file" of viral DNA fragments. But if you look in more detail, it turns out that several different types and classes of CRISPR systems are already known, which have different and peculiar biology. How do these classes differ from each other?

– In principle, different CRISPR systems are, as they say in English, different ways to skin a cat. All of them are the result of an endless war between bacteria and their parasites. What unites CRISPR systems is that they are adaptive immunity systems. Adaptivity in this case means that the system can capture a fragment of the genetic information of the parasite (it can be not only a virus, but also a plasmid or some other mobile genetic element) and embed it into a CRISPR cassette, which is part of the bacterial genome. And such a well-acquired fragment will work as a molecular memory, allowing the descendants of the bacterium to detect and prevent infections by the "remembered" parasite.

Even such a superficial description implies the existence of two stages of action of CRISPR systems: the stage of adaptation, that is, the acquisition of "molecular memory", and the stage of its application - what you do with this memory, how you use it to fight parasites. So, the only thing that unites different types of CRISPR systems is the adaptation module. It consists of two proteins Cas1 and Cas2, which are responsible for capturing foreign genetic information and embedding it in a CRISPR cassette. These proteins are encoded in all CRISPR systems.

But when the first task is solved, when the necessary information is already in the CRISPR cassette, you can destroy the parasite in a variety of ways. In the course of evolution, there have been several cases of combining the same adaptation module with different interference modules – "machines" for the destruction of alien genetic information. As a result, different classes and types of CRISPR systems have emerged. One of the co-authors of our article, bioinformatician Evgeny Kunin, is actively engaged in the reconstruction of their natural history.

The classification of CRISPR systems is based on differences in interference modules - performers, or executors. These performers must be able to do a lot of things: first they must physically interact with the "memory" stored in the CRISPR cassette in the form of RNA, then bring it to the target, try it on and, in case of coincidence, destroy this target. There are two classes of CRISPR systems: in Class II, a single but very large protein does all the executions, in Class I systems these functions are divided between several separate, different proteins, each of which plays its own role. Only second-class systems, where a single multi-cell protein does everything, have become widely known. Simply because, from the point of view of practical applications, it is much easier to work with one protein than with a dozen.

Within each class, several types of performers are distinguished, which are united according to the principle of common origin. There is, for example, the Cas9 protein, which is now widely used in genome editing, and there is the Cpf1 protein, which does about the same thing and has about the same size, but is arranged differently and is not related to Cas9 proteins. The situation is about the same as with the wings of butterflies, birds and bats: they do the same thing, but have completely different origins.

The new system, which we, together with Evgeny Kunin and Feng Zhang, found and describe in our article, is similar to CRISPR/Cas9 in its work, in ideology, but is not related to either it or Cpf1. The target is found and destroyed by a new, previously unknown protein called C2c2. And since he is unrelated, he does everything differently. This does not mean that it will necessarily work better than known proteins, but it may have advantages important for biotechnology and medicine.

Systems of the first class (type I, III and IV) are multi–unit. From the point of view of biology, in my opinion, they are much more interesting (because they are complex, complexity is always exciting, it is interesting to understand it), but from the point of view of practical application, of course, it is harder to work with them. Therefore, biotechnologists are avoiding them for now. Although, in principle, due to the separation of functions, the entire process of searching, analyzing and destroying a target in such systems becomes much more accurate.

– What, it would seem, is just very good for genomic editing?

– Yes, well, but these systems are very complex and cumbersome. After all, in order to edit the genome, you need to pack all the excutatory tools into some kind of delivery vehicle to the eukaryotic cell. For example, in the adenovirus. And the "carrying capacity" of the adenovirus, in fact, the length of genes that can be packed into the viral envelope, unfortunately, is very limited.

Scheme of operation of the CRISPR-Cas type II system (Wikimedia Commons)

CRISPR-Cas Type III system (Wikimedia Commons)

Nevertheless, there are some very interesting multi-unit systems. For example, type III systems that act as targets not on DNA, but on RNA. Why is this necessary? The fact is that many viruses act very cunningly – when they enter a cell, they first embed their genome into the genome of the cell, become part of it. And they sit there, quieter than water, lower than grass, presenting the cell with the opportunity to divide, and therefore increase the number of viral genomes. If you have a memory of such a virus in your CRISPR cassette, then the system will find it and, of course, destroy it. And your own cellular DNA will get "distributed" along with the viral one, the cell will die. Therefore, "smart" type III systems are looking for a target not in the genome itself, but only among working copies of genes, those that produce RNA. As long as the virus in the genome sits quietly and does nothing, its genes are inactive and the immune system does not see it and does not touch it. As soon as the virus begins to multiply, synthesize viral RNA (on the basis of which proteins and viral particles will then turn out), the cell "understands" that something very bad is happening to it and it's time to act. The type III CRISPR system activates and destroys the virus, however, at the same time, destroying the host cell.

– The system based on the C2c2 protein that you have just found works in about the same way. But before we talk about it, tell us how the search would be organized, how did you learn to search for new CRISPR systems in general? As far as I understand, a special search engine was made for this, which is described in your previous article, right?

– Yes, it was made by Seryozha Shmakov, who first studied at Baumanka, then worked at Microsoft, then came to the Yandex Data Analysis Evening School, where he learned how to work with genomic data. And then I convinced him to go to Skoltech because he got bored at Microsoft.

What he did is quite easy to explain based on what we have already talked about. All CRISPR systems must have adaptive and interference modules. And the first of them is actually the same for all known systems – it consists of homologous proteins Cas1 and Cas2. The search engine that Seryozha wrote does three things. First, he finds in the databases of genomic sequences all the pieces of DNA encoding proteins similar to Cas1. Secondly, the search engine looks at sequences located nearby and throws out already known CRISPR effectors, they are not interesting to us. And, thirdly, the search engine looks for any large genes in the sequences adjacent to the predicted Cas1 genes - simply because we believe that a protein that could cope with all the functions of the CRISPR effector should be large enough. That's all, in three steps we get a list of candidates for the "new Cas9".

After the list of candidates is ready, the moment comes to open the cards. If we really learned how to properly look for CRISPR effectors, these multi-cell proteins, then somewhere near them there should be cassettes in which the memory of viruses is stored. We did not specifically search for these cassettes, their presence is embedded in the search algorithm, but they should be there simply based on the logic of the system. And indeed, they are found there, i.e., we are on the right track.

– And it would be possible to act the other way around – to look for cassettes and find effectors.

– Yes, such a search engine is also made now. So, among the list of candidates, we found three completely new groups of effectors. They were named candidates-1, 2, and 3 (C2c2 is simply "Class 2 candidate 2"). In our first article in Molecular Cell, we confirmed that C2c1 really works. That when transferred to E. coli, such a system, predicted by us in the genome of a completely different bacterium, really protects against infection with foreign DNA, and the C2c1 effector itself really "cracks" the DNA corresponding to the sections of the CRISPR cassette in a test tube, and so on.

The new protein is quite interesting from a biotechnological point of view, as a replacement for Cas9. For two reasons: unlike Cas9, it turned out to have a different kind of "raskus" of the target DNA, more suitable for genome editing. Secondly, it has a different specificity to additional nucleotides, that is, to PAM. This specificity significantly limits the applicability of Cas9 for genome editing.

– What is PAM?

PAM (protospacer adjacent motive, "the motif adjacent to the protospacer") is one of the most interesting pieces in the whole CRISPR kitchen. Interesting first of all because it could be opened on the tip of a pen, without any experiments. For the first time, Francisco Mojica drew attention to its existence – a stunningly intelligent, but completely untwisted person in this area. By the way, he proposed the term "CRISPR".

Look, if bacterial immunity works the way I told you, then you should guess for yourself what PAM is. So, you have a cassette with fragments of viral DNA (spacers), RNA is synthesized from the cassette, RNA binds to the effector and if the CRISPR RNA spacer matches the target, such a target will be destroyed. It is clear that the first thing that a coincidence will happen to is the cassette itself in the genome from which the RNA was synthesized. It turns out that the bacterium will be forced to kill itself? It seems to be, but it doesn't happen, and here's why.

The fact is that in order for the effector protein to start cutting DNA, it is not enough to completely match the guide RNA with the target. One more condition must be met: a special short sequence must be attached to the target, which tells the protein that it is not its own genome. There is a repeat in the genomic cassette next to the spacer, which the effector protein feels as "its own". If the same sequence exists in the target, it will not be cut. These are only a few nucleotides, but they are enough to save the target. In order for the target to be cut, some completely different sequence must fit to it, as different as possible from what the cassette has. Such a sequence is RAM. It provides "friend-foe" recognition.

The ability to recognize PAM is built into the effector protein. It is clear that for biotechnology, this means that, generally speaking, we cannot edit any parts of genes. Only those where there is a suitable PAM, and this greatly shackles the hands.

– And therefore the necessary new effectors that you are looking for. Tell us about the C2c2 nuclease that you found in the new article. What is it, how does it work?

– This is an effector that acts exclusively at the RNA level. Unlike Type III systems, which are RNA-dependent but cut DNA, C2c2 acts only on RNA. Such a system has its advantages. You wrote about it in the news, and we also mentioned it in the article – that it could potentially be used to act on cellular RNAs or something similar. But, I'm afraid, in fact, it will not have practical significance, unfortunately. But from the point of view of fundamental science, this system is very beautiful.

After all, we are talking about the immunity of unicellular organisms, which is fundamentally different from our own immunity. In unicellular, in a clonal population (where almost all individuals are genetically identical to each other), everything is arranged roughly as in Stalinist society: the life of each individual cell is nothing. If there is a viral infection, it is more profitable for the population to let infected cells die, but to prevent the spread of the virus through the population, than to somehow try to save an individual infected cell.

– And how does C2c2 kill the affected cells?

– The military has such a concept – collateral damage, "collateral damage". This is when, for example, the Americans bombed Belgrade, and got not only where they were aiming, but also at the Chinese embassy at the same time. Here it works about the same way. The C2c2 effector recognizes the target RNA and, like Cas9, cuts it - but, unlike the latter, it cuts not in the same place where it recognizes, but in another place of the RNA.

We started to study this and it turned out that in C2c2, target recognition does not lead to an accurate cut, but is simply a signal that a viral infection has appeared in the cell. And as soon as such a signal is received, the cell decides to take its own life, and the effector begins to destroy all the RNA available in the cell. The infected bacterium, of course, dies. Along with the virus. It is clear that this is not very suitable for genomic editing at the level of individual genes.

– Yes, but it can be used to destroy cells expressing "wrong" RNAs. For example, alpha-fetoprotein, characteristic of many tumors, or other cancer markers?

– Yes, you can think about it – if, for example, it was possible to find such a transcript, an RNA molecule that is characteristic only of a tumor. In principle, C2c2 itself is very accurate, we checked it – one mismatch is enough to prevent its false triggering. So, in principle, what you are suggesting is possible, but not as a system for suppressing the activity of specific genes by the type of RNA interference, namely, as induced, programmed cell death. Another paradigm.

– As I understand it, you are not too keen on the applicability of what you are studying in medicine?

– I'm definitely not. But one of our co-authors, Feng Zhang, is quite a yes. He wants to cure people of cancer, control the brain, etc. I am interested in all this, first of all, because it allows me to study new biology on the wave of CRISPR madness, to shift the boundary of knowledge. As for the search for new effectors as a replacement for Cas9 in biotechnology, they already exist and there are more and more of them.

For example, Cpf1 [found by Zhang and co-authors] is being actively used. The search engine we just talked about has given us another 5-6 completely new candidates that still need to be investigated and investigated. You can go the other way and improve the already known Cas9 proteins, as was done recently – when its accuracy was increased by replacing several amino acids. Or when they reduced the size of a natural protein by removing 20% of amino acids from it. Can you imagine an article in Nature about how a deletion was made in a gene? Coursework, at best. On the one hand, it's funny, on the other hand, there is much more space in the adenovirus, with which delivery to cells is made. From the point of view of biotechnology, this is very important.

– And, nevertheless, you posted your article in bioArxive before publication. Obviously in order to guarantee priority for subsequent applications?

– Yes, and for that too. And we got a patent for this case – jointly with MIT, Rutgers University and Skoltech.

– Tell us about the conference in Israel, from which you have just returned. I understand that everyone was there: the discoverers of CRISPR, and those companies that are trying to turn biotechnology into therapy. What is happening in the "CRISPR business" now? Well, besides the fact that there's crazy money spinning around?

– There was no question of money, because this is a purely fundamental non-profit conference. Which, by the way, was initially invented and supported not by companies like EDITAS (an American startup that is going to be one of the first to use CRISPR/Cas9 technology for therapy – note N+1), but by the American Air Force.

Nevertheless, it is now quite obvious from communication with both "fundamental" and "commercial" people that the main problem of technology is the bioinformatic prediction of targets for guide RNAs when editing certain genes of humans, animals and plants. The fact is that accurate recognition is not limited only to the correspondence of the target and RNA, the presence of PAM-a in the gene, or the presence of erroneous (off-target) activity in the effector. To find a good guide RNA for genome editing, it is necessary to take into account both the state of chromatin, the position of nucleosomes and transcriptional activity at a given location of the genome and, most likely, a lot of other factors. Now everyone is working to come up with a computer model for predicting the right targets and guide RNAs.

– Do they want to turn this prediction into a service?

– Yes, to the service for the end user. The main problem in this whole business is not only and not so much to learn how to make mutations effectively, but to ensure that they are correct. Because after making breaks, different events can occur – there may be a deletion in this place of the gene (the gene will be broken), or, conversely, a replacement with a pre-selected "correct" copy of the gene will occur. It is difficult to predict what scenario this process will follow now. It depends on a lot of factors: the guiding RNA, the type of effector, the state of the cell, and so on. There will be a lot of data – we will analyze it. Sergei Shmakov recently organized a hackathon in Moscow on this task, trying to find the factors that affect it.

– Machine learning?

– Exactly, machine learning. The effectiveness of the process can be influenced by the context of the target (local or global), kinetic parameters, whatever. There is experimental data, and it is possible to identify these parameters, and then make predictions and check them by adjusting the system.

– But it follows from all this, apparently, that the recently initiated patent dispute over the ownership of CRISPR technology will not affect anything anymore?

– Yes, it's like with restrictases – no one has made much money for them. By itself, the patent–protected Cas9 is only a small part of the story. It's easy to get around a patent for it, that's not the problem. The main intellectual property will be enclosed in specific guide RNAs that will be used to treat specific diseases. No one has reached this stage yet, so in this sense, the main division has not yet begun.

– Which human diseases will be the first in CRISPR therapy?

– Everyone wants to treat retinal diseases. And Shinya Yamanaka (Nobel laureate, discoverer of the method of producing induced stem cells – approx. N+1) is doing this, and EDITAS. It is easier to introduce genetic constructs into the eye. Plus, liver and muscles: muscle degeneration is just a good option for a test disease, since you can simply inject a muscle and even a small percentage of cured cells will play a role for therapy. And, of course, blood, – chimeric T-cell receptors want to do everything.

– And what about editing the genome of embryos, interfering with the human sexual line?

– Jennifer Dudna (one of the authors of the genome editing technology – note N+1) made a big report on this. The conclusion, in general, is that even she (despite the fact that she has always been sharply opposed) recognizes that there is no way to stop the spread of technology. Whatever we think about it. By the way, she said that in the near future, California will allow work on the modification of the sexual line.

– I just can't understand why interference in the sexual line should be feared? Can you explain to me why it's dangerous?

– To be honest, I don't know either. It's like with GMOs. The only thing to be afraid of is ignorance, it seems to me.

Portal "Eternal youth" http://vechnayamolodost.ru  06.06.2016