At A Glance
Name: Nikolaus Rajewsky
Position: Assistant professor, biology/mathematics, New York University
Background: Research professor, Rockefeller University — 2002; Postdoc, Rockefeller University — 1999-2002; Postdoc, Rutgers University — 1998-1999; PhD, theoretical physics, University of Cologne — 1997
Although he didn’t even begin to work in the biological sciences until his second postdoc, Nikolaus Rajewsky has embraced his work in this field. Most recently, he co-authored a paper in Nature dealing with the identification of a novel microRNA that turns out to be associated with diabetes.
Rajewsky spoke with RNAi News this week about his work.
How did you get involved with microRNAs?
I just read a lot of papers, and then I noticed a couple of papers about microRNAs, which I thought were very interesting. From an abstract point of view, they are especially interesting because I was working on transcription before, and if you have a transcription factor, without doing experiments, you have very little chance of guessing what this transcription factor is going to recognize and bind to; the cis-acting elements are difficult to guess if I just pick a transcription factor. But it appeared that [with] these little microRNAs, you had a chance of guessing what they were binding to and regulating just given sequence information. I thought that was very interesting.
And, of course, the papers which had come out about C. elegans … [and] developmental timing, they were of interest to me because I was working already in developmental biology quite a bit — I was interested in the body-patterning system in flies. I was thinking about developmental biology, and therefore I became interested in this.
But I remember that when I wrote my job applications [after my postdoctoral work], I didn’t actually [mention] microRNAs because I thought it was too exotic.
But then, when I came to NYU, I had a couple of months to basically do whatever I wanted to do. In these days there were quite a few papers coming out about microRNAs showing that you find them basically everywhere. I was literally reading one of the papers when I got a phone call from Markus Stoffel [at Rockefeller] asking me if I would like to participate in [a microRNA] project with him. Then the whole thing started to get serious.
Given that you were at Rockefeller and interested in microRNAs, did you ever run into Tom Tuschl?
Actually, no. I had just heard then that he was hired, but I never ran into him at Rockefeller. Really, it all started when I was at NYU … and I got the phone call from Markus Stoffel.
Can you talk a little bit about going from reading that paper and getting the phone call to what you are doing now?
Basically, Markus called me and asked if I could compute some possible targets for some interesting microRNAs I have found. I looked around and there was no paper about how to predict targets, so I started to think about it and work on it. I started by analyzing the very few known targets in C. elegans and downloading the sequences and playing around with them, then asking a friend [at NYU] Nick Socci to join me on this project.
We had a very exciting summer just trying to play around with these data. It was pretty clear that some basic work had to be done first to define some algorithm for target finding before I could even start working with Markus Stoffel.
I am grateful to Tom Tuschl because he gave me a whole set of microRNAs in flies which weren’t published then so I could experiment with those, and that’s what led to my paper in Developmental Biology [published earlier this year].
After that, I continued to play around with Stoffel’s data, and that came to this [Nature] paper.
This Developmental Biology paper, could you touch on it a bit?
Basically, what we did was try to define a set of target sites for microRNAs [that] were reasonably well established. Then we tried a whole variety of methods, and screened the parameter sets for each method, trying to arrive at the method with the parameter sets that would best recover the known targets when comparing to random sequences.
It’s really like a physicist’s approach; it’s really scanning systematically a large number of parameters and scoring each method, and thus finding the method that is best in recovering the sites when compared to random sites.
What we found is our method, which by now has changed a bit. What came out of the data is that the most decisive step in target recognition seems to be a string of consecutive base pairings of the microRNA to the target site.
The very interesting point was that we could make a model, a very crude [one], but still a model for microRNA-target recognition [that showed] there is a kinetic step [that] is decisive for target-site recognition, and then a thermodynamic step [that] is just stabilization of the microRNA having recognized the target.
The idea [is] that you need consecutive base pairings [that] are energetically favorable such that the microRNA can zip up rapidly enough with the target site to overcome thermal diffusion. You have to imagine that the microRNA is diffusing in the cytoplasm searching for targets, and has only a certain time scale to recognize a target. Our interpretation is that these consecutive base pairings help to nucleate the target recognition zipping up rapidly enough to overcome thermal diffusion.
The rest of the microRNA is somewhat more loosely bound to the target site, and the overall free energy of this binding is much less predictable for target sites, but it still appears that it has to have a certain minimum quality. So you can’t go above a certain free energy for good target sites. … When we submitted this paper, there was no paper about target recognition around, so we were very happy to have found a method [that] allowed an interpretation in terms of a model. …
Then I continued my interaction with Markus Stoffel.
Can you give some insight into the work with Markus [and the resultant Nature] paper?
[Stoffel and his colleagues] cloned and sequenced microRNA [that are specifically expressed] in pancreatic islet cells. Then they found that a certain microRNA, microRNA 375 — a novel microRNA, by the way — would modulate insulin secretion. In fact, it would suppress glucose-induced insulin secretion. That was, of course, extremely exciting, but to understand what this microRNA is doing you have to say something about its regulatory targets. That’s the point where he asked me to help predict some targets [so that] he could validate them.
What happened was a very intense collaboration between him and Matt Poy, the [paper’s] first author, and me, where a large number of lists [of possible targets] were going back and forth to see if the results made sense and if they liked the lists or not.
Then I changed the algorithm and changed the datasets. Just to define a good set of 3’ UTRs in mammals is [not a] trivial enterprise. I think what really made this work was very functional and intense communication between us.
In the end … they tried to see from a physiologically point of view where in the pathway is a microRNA really acting. They found it was acting at the level of exocytosis. When they looked at my list of predicted targets, which was a relatively large list of roughly 65 predicted targets, they picked a few that made sense in terms of their functional annotation [and] were consistent with this picture of the microRNA doing something at the level of exocytosis. They tested those and they turned out to be valid. …
This is the first paper [that] defines a good biological function for a mammalian microRNA, and I think that should be said. Also, it’s the first paper where a computationally predicted mammalian target is really validated in a non-trivial way.
[Additionally, the paper] showed that when you knock down myotrophin, it mimics the effect of over-expressing microRNA 375. That’s very pretty because it puts myotrophin into the exocytosis pathway, which was not known before — myotrophin had not been associated with insulin secretion at all.
The algorithm and dataset development is constantly evolving, and right now I have a new algorithm [that] is not published, where I think I can understand much better why myotrophin turned out to be a microRNA target involved in insulin secretion.
My point is that things in the microRNA target game are, in my opinion, in a very early stage. I don’t believe that one algorithm is better than another, actually. We cannot say that because the algorithms are all based on the few known microRNA targets for a few microRNAs.
This is all going to change. It’s getting more sophisticated. I’m thinking that the more targets we find are going to boost the efficacy of the algorithms. Right now, it’s the very tip of the iceberg. … I think there is a chance these algorithms are going to work quite reliability, but it will take many years for that to happen.