Visiting “Noncodarnia”

Long noncoding RNAs (lncRNAs) are everywhere these days. In the past few months, major reviews have appeared in Science, Genetics, and Nature Structural & Molecular Biology. In April, Nature profiled the growing job prospects in lncRNA research, noting, “Enthusiasm for lncRNA has replaced much of the science community's scepticism” (1).

A decade ago, things looked very different. Although researchers were aware of a few biologically important long noncoding RNAs, such as Xist and H19, these were largely discounted as special cases. In fact, so ingrained was this opinion that John Mattick, an lncRNA pioneer then at the University of Queensland, was pretty much flying solo in 2001 when he proposed that regulatory RNAs, not proteins, “form the primary control architecture that underpins eukaryotic differentiation and development” (2).

“It seemed to me, and it still does, that people have missed the point,” says Mattick, who is now the Executive Director of the Garvan Institute of Medical Research in Sydney, Australia. “Most of the assumptions that we operate on in molecular biology derive from the initial assumption that most genetic information is transacted by proteins. And while that's largely true in bacteria, it's not true for humans.”

In bacteria, RNA largely exits in its most well-known, well-characterized forms: tRNA, rRNA, and mRNA. But according to Mattick, in humans and other higher eukaryotes, most RNAs orchestrate very specific epigenetic hierarchies that control gene expression in four dimensions. “They're actually the computational engine of multicellular development.”

And a very malleable, robust engine at that. A single segment of the genome may contain layer upon layer of control sequences, thanks to a combination of epigenetic signals, sense and antisense transcription, and alternative splicing. “Each locus, each region of the genome is unzipped—transcribed, that is—in wildly different patterns. It's just extraordinary how densely packed our genome is with information,” explains Mattick.

Today, the broader research community seems finally to have caught on, awarding more research dollars than ever before to efforts focused on characterizing key players within the RNA world.

Still, it's no surprise that the researchers who choose to pursue these curious molecular RNA creatures seem to consider themselves explorers within a strange new world—a world Harvard University geneticist John Rinn calls “Noncodarnia.”

Through the wardrobe

Although Rinn (whose Twitter handle is @noncodarnia) has been pursing noncoding RNAs since graduate school, that's not what he wanted to do when he first enrolled at Yale. “I was certain to be a crystallographer,” he recalls. But during an “amazing rotation” through Michael Snyder's lab, he discovered microarrays and was hooked, studying the transcriptional output first of chromosome 22 and, eventually, the entire genome.

That output included noncoding RNAs, a topic Rinn wanted to pursue as a postdoc. Most potential advisors he approached, though, were unconvinced. Then he met dermatology researcher Howard Chang at Stanford University, who was just setting up his own lab and was willing to take on an unconventional project.

“A lot of people at the time didn't want a postdoc to study a lot of bogus hocus-pocus,” says Rinn. However, the gamble paid off. In 2007, Rinn and Chang discovered HOTAIR, an lncRNA from the human HOX locus that also binds the epigenetic regulator Polycomb Repressive Complex 2 (PRC2). Researchers had previously suspected that noncoding RNAs could play a role in epigenetics, but HOTAIR was the first RNA molecule shown to actually carry out such a function.

“RNA has always been the deviant molecule,” notes the 37-year-old Rinn. “It always will do something you think something couldn't do.”

RNA enzymes? Check. Telomere maintenance? Check. RNA interference? Check, again.

But unlike those other RNAs, lncRNAs don't have a single, defined function. “We're calling long noncoding RNAs a class, when actually the only definition is that they are longer than 200 bp,” says Ana Marques, a Research Fellow at the University of Oxford who uses evolutionary approaches to understand lncRNA function. As a consequence, working out the function of novel lncRNAs is very difficult.

“You may have to clone 15 lncRNAs before you're going to find one that has an effect that you might have been hypothesizing,” says Rinn.

Homing in on function

Researchers have managed to decode the functions of a few lncRNAs—perhaps 1% of all human noncoding transcripts, Marques estimates. Some, like Xist and the newly discovered Braveheart, which are implicated in X-inactivation and cardiac development, respectively, are nuclear transcripts that, like HOTAIR, bind Polycomb-group proteins. Another class of lncRNAs, called 1/2-sbsRNAs, are cytoplasmic molecules that mediate RNA decay. And still others, like lincRNA-RoR, act as molecular sponges, soaking up and neutralizing microRNAs.

“There is no unifying umbrella like for microRNAs,” says Leonard Lipovich, a lncRNA expert at Wayne State University. In a recent review on lncRNAs, Jeannie Lee, a Howard Hughes Medical Institute Investigator at Harvard Medical School who studies X-inactivation, includes a figure illustrating some 16 distinct functions for lncRNAs (3).

“It's like you're going through the wardrobe into this crazy, mysterious land, and it's both exciting and scary at the same time,” says Rinn. “You're seeing all these mysterious creatures you've never seen before and yet, you don't know how to navigate it, you don't know what to do.”

Rinn's lab focuses on RNA mapping and functional analysis of long intergenic noncoding RNAs (lincRNAs). They use next-gen DNA sequencing to chart the former; for the latter, it's guilt-by-association: watching for coding transcripts whose abundance rises and falls with that of a specific lincRNA. A match provides evidence that the lncRNA and protein might function in the same pathway, a hypothesis that then must be tested.

The process is high-tech, but crude. “It's almost like looking at Noncodarnia thru an Atari rather than an Xbox,” says Rinn, who tends to pepper his conversation with pop-culture references. Unlike protein researchers, “We're in a low-res form of the field.”

That's in part because there are so few tools available specifically for studying lncRNA. Rinn says the one constant in his career has been the need to use what he calls the “the MacGyver approach,” creating solutions from tools that weren't necessarily designed to address the problem at hand.

In 2008, for instance, his team was having trouble seeing low-abundance lncRNAs against a background of more abundant protein-coding transcripts. So, they used DNA microarrays to enrich noncoding transcripts first, gaining three orders of magnitude in sequencing depth. More recently, his team has focused on informatics, working out methods to filter, analyze, assemble, and disseminate lncRNA data to the broader scientific community.

“What do you do when you're flying blind? You try and make a flashlight,” he says.

A pitch for staying old-school

Building flashlights is a common task in Noncodarnia. Jeannie Lee has had to do that in her lab as well, developing a method to survey the landscape of protein-complexed RNAs called RIP-seq, which she used in 2010 to identify more than 9,000 PRC2-associated transcripts.

Yet like Rinn, Lee wasn't looking to study lncRNAs when she entered the field in the early 1990s. She was studying X-inactivation as a postdoc with Rudolf Jaenisch at MIT around the time that Xist was identified.

“I remember very clearly going to a national conference at which the discovery was presented. And I was blown away by the fact that a long noncoding RNA that's coating the X chromosomes could somehow be involved.”

Once she established her own lab at Massachusetts General Hospital, Lee continued to focus on X-inactivation and very quickly identified an lncRNA antisense to Xist. Called Tsix, the newly identified molecule regulates Xist and together with other lncRNAs enables dosage compensation in XX females.

Nearly a decade later, Lee's lab found a binding site within Xist for the Polycomb complex PRC2, directing PRC2 to the soon-to-be-inactive X chromosome, an observation that ties lncRNAs to X-inactivation and provides a possible solution to an epigenetic riddle.

Epigenetic regulation is locus-specific— some genes are turned on while others are turned off. Yet epigenetic factors show little if any sequence-specificity. Lee suggests that lncRNAs can bridge that gap, because they can fold up to bind proteins but are also inherently sequence-targeted.

“When you're studying epigenetics and you need locus specificity,” she says, “you have to look to long noncoding RNAs.”

Today's researchers have RNA-seq and RNAi to tackle that job, not to mention databases overflowing with lncRNA sequence data they can search at the click of a mouse. But when Lee started her lab in 1997, there was practically nothing upon which to build. “Imagine to yourself, you've set out on a career to figure out how X-inactivation works, and there's only one molecule,” she says, laughing.

Lee had to make do with good old-fashioned genetics, making knockouts in embryonic stem cells and looking for changes in phenotype. As a postdoc, she used random deletions to identify a 450 kb sequence of the X-inactivation center that is both necessary and sufficient for X-inactivation, and it was in this sequence that her lab found Tsix.

Today, a popular choice for querying lncRNA function is genetic knockdown, using either siRNAs or shRNAs. Lee's lab uses these approaches, too, but also remains true to the technique that served her so well in the past. “I want to make a pitch here for the importance of making traditional knockouts to demonstrate function,” she says. “A lot of these knockdown approaches to lncRNA biology yield phenotypes that do not ultimately agree with the knockout,” though both approaches have their place in functional studies, she emphasizes.

Countering resistance

To get a sense of the difficulties researchers face in Noncodarnia, look no further than Case Western University biochemist Saba Valadkhan.

When Valadkhan started her lncRNA project in 2006, there were no deep sequencing or large-scale transcriptomic data available. “We had to rely on a very small number of accidentally discovered long noncoding RNAs to choose one for our studies.”

Not wanting to step on anybody's toes, Valadkhan searched the literature for interesting but “abandoned” lncRNAs. She settled on a bone morphogenic protein-induced transcript first described in 1998.

But working with that transcript, she says, has been a challenge. “We really didn't even know what to worry about and what would be okay.” For instance, when cloning the transcript, what would happen if extra sequences were tacked onto the 5′ or 3′ ends? “They can potentially change the secondary structure of the RNA in these regions... or they can be binding sites for proteins that shouldn't be bound to this RNA.” Knockdowns weren't obvious, either. Where, for instance, should she target the interfering RNAs?

Function was a problem, as was the dearth of knowledgeable colleagues. And so, when the data started rolling in, Valadkhan's team wasn't sure how to deal with it. They didn't know, for instance, how much of a phenotypic response was real as opposed to noise. The only thing they knew for sure was that their RNA seemed to be doing something amazing.

When overexpressed in myoblasts, the transcript caused the cells to “change shape strongly and [start] to grow little processes.” The team had no idea what that meant, but the postdoc doing this work thought the cells kind of looked like neurons. “Lo and behold they were screaming with neuronal marker expression,” Valadkhan says.

This was in 2007, she says, shortly after Shinya Yamanaka introduced the world to induced pluripotent stem cells. Nobody doubts the power of transcription factors to alter cell fate, yet Valadkhan's discovery that noncoding RNAs evidently do the same, remains unpublished.

“The reviewers hated it,” she says. “They said this must be an artifact.”

Reviewers asked for control after control. They demanded knockout mice. They even questioned Valadkhan, whose background is in “hardcore RNA chemistry,” on her ability to culture cells.

Six years later the paper is once again under review. “Again they have asked for quite a number of things,” she says, “but at least the tone of the reviewers has become more and more positive with every submission. I think it's partly because we are adding a lot more controls and data,” she allows. “But partly also because people are appreciating that these RNAs can really do these kinds of things.”

For too long, Mattick says, the broad molecular biology community has acted as if it had the conceptual framework of cellular genetic regulation more or less figured out, thanks to promoters, enhancers, and transcription factors. “They decided they understood the system 50 years ago, after just a decade or so of initial work, and haven't really changed their minds since.”

That view seems finally to be changing, Mattick says.

It looks to be a fine day in Noncodarnia.