At A Glance
Name: Todd Lowe
Position: Assistant professor, biomolecular engineering, University of California, Santa Cruz
Background: Postdoc, Stanford University — 1999-2001; PhD, molecular genetics, Washington University, St. Louis — 1999; Teaching assistant, Washington University, St. Louis — 1994; Computer specialist, National Center for Biotechnology Information — 1992-1993; BA, biology, Williams College — 1992
How did you get involved with the field of non-coding RNAs?
When I was in graduate school in Sean Eddy’s lab at Washington University in St. Louis I first started working on scanning for [transfer] RNAs. They needed an improved tRNA scanner because the one that they had was expected to give as many false positives … as true positives … when scanning the human genome. So I started working on that, and the result was called tRNAscan-SE; it’s probably the most used tRNA scanner in the field right now, and it has been since that work in 1997.
From there I moved on to looking for small nucleolar RNAs in the yeast genome, which had just been published. We knew that the functions of these guys implied that there were a lot more unfound genes that were lurking somewhere in the yeast genome. So for my graduate work I wrote another probabilistic gene finder for looking for snoRNAs. [I] identified a bunch of new [ones] and went into the wet lab and disrupted them, and showed that the expected phenotype resulted when removing the snoRNAs. This was sort of the first large class of RNAs that have these antisense regions in them — now microRNAs have stolen the spotlight, but snoRNAs were really the first ones that [made] people realize there were a lot of RNAs lurking around genomes that protein-encoding gene finders were not able to find at all.
The papers that resulted from that [work] helped advertise that non-coding RNAs are worth paying attention to. Before then, people pretty much only cared about protein-encoding genes.
That’s how I got into the RNA field.
Can you give an overview of … what snoRNAs are and their role?
SnoRNAs are involved in processing, primarily of ribosomal RNA, but they do actually modify other RNAs in specific circumstances like splicesomal RNAs. A couple of them are involved in chopping up the ribosomal RNA transcript — originally, the small and large subunit are transcribing all in one piece in eukaryotes, and there are a couple snoRNAs that will process that and are essential for life. The rest, the majority, are involved in guiding the location of two types of modifications.
In humans there are probably over 150 or 200 snoRNAs, and in yeast, for example, there are dozens and dozens of these guys. Through the 80s, people didn’t know what these things were doing — they were looking for splicesomal RNAs and they found all these other RNAs floating around.
There are two classes of snoRNAs: there’s the C/D box snoRNAs, and it turns out those are involved in adding a methyl group on the ribose moiety of, primarily, ribosomal RNAs. The other major class is the H/ACA box snoRNAs, and they are involved in changing a uridine to pseudouridine. These are all post-transcriptional modifications of ribosomal RNA that have to occur after the pre-RNA has been transcribed and before it can be assembled into a ribosome.
The phenomenal thing about these modifications is, at least for the ribose methylation, that we don’t have an exact function for [them] yet. There are some hints that perhaps they might be involved in helping get the RNA folded in the correct pathway and help speed folding, but we don’t really know. In yeast, all the snoRNAs that guide ribose methylation that I disrupted and other people disrupted were non-essential. So you can remove the individual modifications, and yet there were no obvious growth defects. That’s not exactly the case with the pseudouridine modifications; with pseudouridines, if you start removing some key ones then you do get growth phenotypes.
The other intriguing thing … [is that] in yeast, even though there are 55 sites to a human’s 100 sites of ribose methylation, two-thirds of the sites in yeast are precisely in the same structural position in the ribosomal RNA as they are in humans. So even though the individual sites are not absolutely essential, they are very, very highly conserved, so we do know they offer some sort of evolutionary advantage.
But the pseudouridines often occur in the core of the pathway where the mRNA gets threaded through the ribosome, and they more clearly modify interactions with other molecules.
Does the work you are doing now touch on microRNAs?
Right now, we are not [working in that field]. There are a lot of other groups that, because they have started cloning [miRNAs] using experimental methods, have jumped on these gene predictors for microRNAs. And so it’s basically a field where I, at the time [of my postdoc] had moved on [from].
We are still doing work in RNAs, but the field of microRNAs has heated up so quickly and gotten so much attention that … it’s not something that I feel like I’d want to get into because there’s already quite a bit of brain power going into it.
Where is your work headed? Where is this all going?
We apply microarrays to look at intergenic regions of genomes … but we’re also improving our computational methods. The field is really concerned with where the RNAs are, and they pay less attention to what they are doing. I think more effort needs to go into this.
With the comparative genomics we’ve got, all the non-coding RNAs, all the things that are transcribed and are conserved, are going to be obvious in the coming years. You’re not even going to need sophisticated computational methods to find them. Figuring out what they do is really the next big thing the field should be most concerned with. That is something that the experimentalists need to have a large amount of input on.
I think the computational folks by predicting interactions, for example, in microRNA base pairing to potential targets, that’s the best kind of contribution they can make towards figuring out what some of these RNAs are doing — by predicting interactions and then having the experimental folks go into it. I think an integrated approach, where you’re doing computational work as well as experimental work, is the most effective method at getting what these things do.
Another thing that I’m interested in is the fact that, right now, if you have an anonymous sequence that you’ve cloned or you’ve got a whole genome that you’ve just sequenced, most of the genomes are very poorly annotated for RNAs. Basically, at best you usually have the splicesomal RNAs and the tRNAs and the ribosomal RNAs annotated, and you only have the tRNAs annotated because tRNAscan-SE is such an easy tool to use. For these other kinds of RNAs, you need RNA specialists in order to look at the genome specifically. There are a lot of programs out there that are either customized for finding one kind of RNA, like microRNAs or snoRNAs or tRNAs; or there’s an excellent resource called Rfam, which is a generalized library of covariance models that allows you to search for classed of RNAs.
But [Rfam] only really gives you the boundaries of the molecule — it doesn’t tell you, for example if it’s a tRNA, or what the tRNA does. You need a customized program that knows where to look in the molecule for specific features. So Rfam is a good attempt, but what our lab is going to be working on in the near future is sort of an ber-RNA classifier [that] integrates all the best aspects of customized RNA gene finders with Rfam. We tentatively agreed to collaborate with the Rfam developers and put together a web-based tool where you can paste in your sequence or genome, and it will run all of the different kind of RNA gene finders there are out there, as well as Rfam. Then [researchers can] best guess whether its an RNA at all, and if it is [they] can find out any additional information like if it is a tRNA or a snoRNA.
Right now, you have to have a fair knowledge of the field to know where to go looking for those tools and then apply them intelligently. Our goal [to put together] an ber-classifier won’t be easy, and it will take several years to develop, but now is the right time to start developing a tool [that] any Joe Biologist can submit a sequence and get an answer with a fair amount of certainty.
Will [the classifier] include microRNAs?
Absolutely. I think Rfam has done a good job at creating microRNA models, and we’re going to basically try to use any sort of freely available programs. There are several scanning for microRNAs, and that will definitely be a part of the overall pipeline for RNA analysis.
The details for the inclusion of specific programs are being worked out.