For most of the past 10 years, Michael Wolfe has been studying the biochemistry of gamma-secretase, painstakingly detailing its structure, function, and mechanism of action. The main culprits of Alzheimer's disease are short amyloid-beta peptides that build up and clump together in the brain. Gamma-secretase works alongside beta-secretase to cleave the amyloid precursor protein into its fatal, truncated cousin. Several years ago, Wolfe began characterizing beta-secretase, but this time with a new tack: determining how alternative RNA splicing affects the enzyme's function. Recently, Wolfe published work that identified alternative splicing events in beta-secretase.
"It's starting to look like most genes in the human genome are alternatively spliced," says Wolfe. A single gene can generate many types of proteins, which lends the relatively small human genome its extreme diversity of expression. That beta-secretase undergoes splicing is not surprising. "It seems to be a key regulatory event, and a way for the cell to control what kinds of proteins are produced from a single gene," Wolfe says.
Alternative splicing has now been found to exist for at least 70 percent of genes. After DNA is transcribed into an mRNA precursor called pre-mRNA, splicing machinery known as the spliceosome steps in. The spliceosome snips out the introns and patches up the strand to form a finished mRNA, which is then translated into a protein. In alternative RNA splicing events, certain exons are skipped, or specific introns left in, so that the final mRNA sequence can vary, producing different splice isoforms of the same protein.
While there's a lot of interesting pathway analysis being done, the tools to look at this kind of variation are, for the most part, in early stages. Most researchers still use RT-PCR to identify new splice variants or to confirm splice microarray results. Biologists have only just begun to scratch the surface of associating different isoforms with unique functions inside the cell. "Looking at diversity of function is still in its infancy. It's very anecdotal," says Benoit Chabot at Quebec's University of Sherbrooke. "People are working on their gene of interest, and they find different isoforms by cloning the genes and [then] try to figure out what this other isoform is doing. Nobody is doing it in a systematic manner right now." For the most part, large-scale tools like splice arrays and up-and-coming tools such as multiplexed PCR and high-throughput sequencing are just beginning to enter the alternative splicing research world.
Protein by protein
Like Wolfe, Brown University's Diane Lipscombe uses splicing as a window into the expression diversity of her protein of interest, voltage-gated calcium channels. Her lab was one of the first to clone these proteins in neurons, where they're important in modulating processes as diverse as gene transcription and neurotransmitter release, and have been linked to epilepsy and migraines. In her current studies, Lipscombe looks at how neuronal-specific factors affect splice variation in the channel and end up fine-tuning its structure and behavior. She typically uses gene databases and comparative sequence analysis to identify a hit, and then goes back in with PCR to see how the splice isoform may be differentially expressed in the tissue of interest. "It's very old-fashioned," she says.
[ pagebreak ]
In work published in 2004, Lipscombe found that a particular isoform of the channel was enriched in nociceptors, neurons that can sense and signal pain. She noted in a recent study that this channel is not only more sensitive to neurotransmitters, but it's also more sensitive to opiates like morphine. "We've long thought that the expression of different splice isoforms probably underlies a lot of the differential effects of drugs in different pathways," she says.
Alternative RNA splicing can vary depending on tissue and stage of development, among other things, and in most circumstances there is a healthy balance between differentially expressed isoforms. Existing side by side in certain percentages, variant transcripts can regulate gene expression by turning protein manufacture on or off. When aberrant splicing occurs — for instance, in some cancers — pathways commit signaling errors which lead to downstream trouble. Biologists are looking for not just whether a particular isoform is there, but whether it's an actual splice variant or just an error in transcription, whether it's tissue-specific, and whether it occurs at a level that's meaningful enough to have any noticeable effect
Peering into pathways
Scott Friedman, chief of the liver disease division at Mount Sinai School of Medicine, knows about devoting a lot of time to a single gene. For more than a decade, he's been studying alternative splicing in a tumor suppressor gene called Krüppel-like factor 6 and how a splice isoform, KLF6 SV1, affects liver cancer. He cloned the gene for KLF6 10 years ago, and after his research turned toward alternative splicing, found that SV1 antagonizes full-length KLF6 suppressor activity. "Transient splicing can be a very subtle fine-tuning mechanism for adjusting cell function," he says, "whereas in cancer, it looks like it's kind of a constitutively on switch that affects cells."
What they found is SV1, "which is really composed of three of the four exons of the tumor suppressor, seems to be drastically up-regulated in many late-stage cancers," says co-author and fellow Mount Sinai scientist John Martignetti. He adds that there seems to be an interaction between the two variants depending on how much of each is present. Friedman and Martignetti recently collaborated on work that fleshes out some of SV1's mechanism of action; using RNAi, they were able to show that knocking down the Ras signaling pathway could decrease SV1 production and inhibit tumor growth. Martignetti's work also looks at prostate and ovarian tumors, and what's going on in signaling dysregulation in these cancers. He's seen up-regulation of SV1 in prostate cancer, where it appears to play a role in many different pathways. "It seems to be involved in changes in proliferation, in metastasis, and even in angiogenesis," he says.
To detect and measure different splice variants, they also use PCR and are trying to develop splice-specific monoclonal antibodies. RNAi using siRNA has been a very effective tool, they say, in that it can be specific enough to distinguish between the tumor suppressor and its oncogenic variant.
[ pagebreak ]
Brown's Lipscombe, however, says the hardest part isn't identifying the variant at the RNA level. "Where we are incredibly limited, where we have a huge hurdle to overcome, is to be able to distinguish at a protein level," she says. "It's a problem. You don't have that much control." As an example, there is a calcium channel variant differing by an exon that encodes only two amino acids, but Lipscombe hasn't yet been able to create antibodies for the two nearly identical isoforms. As a stopgap solution — it's harder, it's more expensive, but it gets the job done — she's created mice that express one or the other.
Despite how difficult it may be, picking up the subtle differences between isoforms is one of the most important steps in pathway analysis. "By having recognized the importance of splicing in [the KLF6] gene, it greatly sensitized our interest in looking for splicing as an explanation for other biologies, using different genes," Friedman says. "Once you recognize it's another level of regulation and you look for it, it's amazing how prevalent these kinds of regulatory pathways are."
Microarrays to market
While tool development is still underway for this field, genome-wide analysis is definitely in swing. In fact, several vendors are hard at work co-opting gene expression arrays to study alternative splicing. Affy's Exon Array has become popular for genome-wide expression analysis, and while ExonHit Therapeutics and Jivan Biologics offer both exon and exon splice junction arrays, the latter have proven to be the truly useful tool in the toolbox. Jivan's Jonathan Bingham says that while arrays that probe exon bodies give you a lot of information, "you're left with the problem, how do those pieces fit together? And you get that answer more directly if you have probes for the splice junctions."
Because most genes are alternatively spliced, says ExonHit's John Jaskowiak, "you need to start taking a higher degree of interrogation of the genome." ExonHit has been around for 10 years, and the company offers array products for both basic research as well as for therapeutic and diagnostic research. It runs an internal research program for Alzheimer's disease, measuring changes across many samples and tissues. "You've got this capability of multiplexing on an array where you don't have as much flexibility to do that much coverage [with] RT-PCR," Jaskowiak says.
To create its arrays, ExonHit mined EST databases and other cDNA genomic information, and then partnered with Affy and Agilent to print them. They offer genome-wide arrays for human and mouse, as well as other arrays specific to druggable-target gene families, such as apoptosis, cytokines, or kinases.
[ pagebreak ]
Jivan also offers genome-wide arrays for human, mouse, rat, fly, and more, as well as targeted arrays for specific applications like oncology or toxicology. "Most of our focus has been on microarrays until more recently," says Bingham, who notes that because data analysis poses such a problem, the company will be making refinements to its software in the future. Most of their initial customers have been academic researchers, but Bingham says he's seen a shift toward using the arrays for diagnostic purposes like studying biomarker signatures, drug response, or toxicity. And with the ability to look across many genes at once with arrays, it's easier to study splicing regulation — a separate field that looks at hundreds of interacting genes and accessory proteins that modulate how and when a gene is spliced.
Not perfect yet
One problem with using microarrays, of course, is that they can't be used to find new variants since only known variants are spotted down. Even if the chip is meant to cover the entire genome, not every splice variant is known for most genes, nor to what extent minor variants play a role in gene regulation. For disease genes, which tend to produce splice variants in precise percentages, finding minor variants is especially important.
"In terms of looking at splicing, we're fairly low tech right now," says Lipscombe at Brown. "What we really need to know is whether a particular mRNA that encodes for a given isoform is present in sufficient quantity that we can say, 'Oh, yes, this must be meaningful.' For sure we could use microarray," she says, but correlating function usually comes back to isolating a variant from one particular population of cells. "Microarray analyses are probably better suited to scanning a whole bunch of different tissues and looking for differential distribution patterns, and so for us, because we know what we're going for — we're just looking at one particular gene — the arrays are not necessarily better than what we're doing right now."
Most of the microarrays are based on EST database information, so they tend to find large exons, Lipscombe adds, ignoring possibly critical drug target regions because they vary by only a few amino acids. "I think there is an increase in databases but there's still a lot of information missing. Right now it's not refined" enough to discover tiny variations or variants in small or unique populations of cells, she says.
Jernej Ule of the MRC Laboratory of Molecular Biology in Cambridge, UK, says he thinks most researchers use exon arrays since it would be too much bioinformatics work to design an array that probes exon junctions. But in coming years, "splice junction microarrays should become standard to replace the current microarrays to characterize disease-related changes and possibly even for diagnostics," he says.
Another limitation to using arrays is that even when a variant is identified, it has to be validated with PCR anyway and tested with functional assays. "I think, in general, the technology is very, very good — and what is important now is to apply it and understand how these changes that are detected are actually being regulated in the cell and which ones are relevant," Ule says. Determining which variants are actually being selected for and which are just errors is an even bigger challenge to the splice research community at large.
[ pagebreak ]
A significant bottleneck for arrays is data analysis. Luiz Penalva, assistant professor at Children's Cancer Research Institute at the University of Texas Health Science Center in San Antonio, says performing the analysis of his experiments, which attempt to identify splice variants in glioblastoma, was the hardest part. "This is one of the major problems with alternative splicing microarrays nowadays," he says. While arrays do come with software analysis packages, none of them is perfect; Penalva says his team, which includes a bioinformaticist, usually has to try many different methods to get trustworthy data.
"One thing that we observe when you get array results, it looks like there are some probes there that are simply not giving you any data, or data that doesn't look correct," Penalva says. "Sometimes … it's better simply to discard this data."
Doug Black, a Howard Hughes investigator at the University of California, Los Angeles, uses a full arsenal of tools to study how splicing is regulated and the role these regulators play in neuronal cell differentiation. In mature neurons, he examines how calcium signaling pathways and chronic depolarization can change splicing. While he incorporates both exon arrays and splice junction arrays into his day-to-day work, he says there are tradeoffs to each. With exon arrays, "[it] is a little bit of a pain and you end up having to validate a lot of what you find through other means," he says. However, when using exon junction arrays, plenty of work has to go into probe design, and coming across new variants isn't possible the way it is with a genome-wide exon array. Still, says Black, the splice junction arrays are more sensitive and more reliable.
As a complementary tool, high-throughput sequencing holds much promise. Black is already using a Solexa sequencer to profile splicing under different conditions. He believes that affordable next-gen sequencing will be able to take RNA splice variant detection to the next level. "It's potentially much more accurate and much easier in the analysis," Black says. In fact, a flurry of studies published recently used RNA sequencing, or RNA-seq, to survey the complete transcriptomes of mice, Arabidopsis, yeast, and human cells. "In practical terms, it's not there yet," Black adds. "You don't get enough individual short reads to sample all the exons that you want to sample. You don't have enough sequence depth without hundreds of thousands of dollars to actually measure splice variants."
[ pagebreak ]
And in the event that you're not wading in funding, Sherbrooke's Benoit Chabot has come up with an affordable alternative to next-gen sequencing. As multiplexed PCR takes off, Benoit has taken advantage of the capacity of newer machines to skip microarrays altogether. Currently he can run 3,000 RT-PCR reactions per day, and he hopes to increase that 10-fold during the next few years. "We're not going to do microarrays," says Chabot. "That means we're not going global as much, but we're going to automate it, make RT-PCR a little bit more high-throughput than what people are using." The approach allows him to look at select splicing events across many tissues and many conditions — experiments that probably wouldn't be affordable using sequencing or microarrays. Whether PCR is used to validate microarrays or as a "replacement for cases where customers are looking at a particular pathway or a particular gene family," says Bingham at Jivan, it will continue to offer up more opportunities in a multiplex setting.
Toward the clinic
As basic research moves ahead with identifying new variants and validating their functions in the cell, clinical research and drug development are already looking closely at how splice variants are dysregulated in disease. Often variants will be expressed alongside one another, but in a healthy percentage. When alternative splicing goes awry, one variant may be more or less expressed, leading to signaling imbalances and disease. It's no surprise that finding drug targets would incorporate the study of splicing.
"Initially, the most interest was coming from the academic world, but what we've seen more recently is that drug companies are starting to do pretty decent-sized studies," says Jivan's Bingham, "because splicing offers potentially more information for biomarkers than gene arrays alone." Splice arrays are used to look at expression signatures for disease state or disease progression, to screen for specific isoform drug targets, or to mark drug response "where splicing changes after a drug is administered," he adds. So far, he notes, the most interest has been in arrays for cell surface and toxicology genes.
In its work developing tools to study Alzheimer's disease, ExonHit began using splicing arrays as a diagnostic to screen patients for clinical trials. By measuring RNA in circulating blood, scientists are able to determine whether someone has Alzheimer's, another form of dementia, or a different disease altogether. Today the only way to accurately pinpoint the disease is post-mortem, so splice signatures have great potential in clinical Alzheimer's research. "The interesting thing about that signature is there are a lot of splicing isoforms that are present," says Jaskowiak at ExonHit. "In many cases, it's the ratio of isoforms that is important."
Harvard's Wolfe hopes that identifying splice variants that cause Alzheimer's will eventually be useful in finding effective drug targets. Today, methods for slowing down the progression of the disease mostly revolve around tweaking gamma-secretase production, but gamma-secretase also plays a central role in the highly conserved Notch signaling pathway. Wolfe's recent work found that beta-secretase also undergoes alternative splicing and could be an alternate drug target. By interfering with its alternate splicing events, one could interfere with beta-secretase function, he says. "Alternative splice isoforms exist. They can vary depending on the cell type, and if we shunt splicing down these alternative pathways, we can lower amyloid production."
[ pagebreak ]
Wolfe has also been studying the impact of splicing on tau, a gene associated with a related, non-Alzheimer's form of dementia. He's found that about half the known mutations in the tau gene change its splicing "and somehow that leads to the self association of tau in clogging up neurons," he says. "What's regulating the alternative splicing of tau? Can we pharmacologically step in and tweak the system?"
Discovering drugs that target one point in a splicing pathway can be difficult considering the number of variables. A recent study out of Chabot's Sherbrooke lab looked at how anticancer drugs affect splicing of Bcl-x to promote apoptosis. "It's known that many types of apoptotic genes are alternatively spliced to produce pro-apoptotic variants or anti-apoptotic variants," Chabot says. To his surprise, he says, no one had ever systematically looked to see if pro-apoptotic drugs actually initiated the apoptotic pathway. In his study, he used 20 drugs on five different cancer cell lines to see how they affected apoptosis. While all the drugs shifted splicing of Bcl-x in the right direction, "it does not do it systematically in all cell lines. Some cell lines respond to it; other cell lines don't, depending on the drug." For other alternatively spliced apoptotic genes, this wasn't the case — for some drugs it went in the right direction and for others it didn't. "It's very complex, and we cannot assume that taking an anticancer drug will always go in the right direction," Chabot adds.
There are a number of ways that alternative splicing events could be manipulated, all of which one day might be applied in the clinic. One approach is to use antisense molecules that would bind to disease-causing transcripts, effectively turning them off. Another might be to make use of nonsense stop codons, which in normal alternative splicing events tell the translation machinery to stop before the full-length protein is complete. In the case of tau, the mRNA forms a tiny loop structure that Wolfe thinks could somehow be made to bind to a small molecule to steer splicing toward one isoform or another. "If there's a structure, it probably has pockets where small molecules can bind, so to the degree we can find structure in the message that regulates splicing, we might identify therapeutic targets to get very specific effects," he says. "There are some attempts at therapeutics, but it's pretty early days."