At A Glance
Name: Bob Goldstein
Position: Assistant professor, department of biology, University of North Carolina at Chapel Hill
Background: Postdoc, University of California, Berkeley — 1996-1999; Postdoc, MRC Laboratory of Molecular Biology — 1992-1996; PhD, zoology, University of Texas at Austin — 1992; BS, biology, Union College — 1988
While his lab primarily studies cell diversity in C. elegans, Bob Goldstein was driven by an interest in what had been happening in the field of RNA interference, along with the efforts of one of his graduate students, to expand the lab’s focus to include research into the gene-silencing mechanism itself. Goldstein recently spoke with RNAi News about how it all happened and where it has led.
How did you first get involved with RNAi?
I started following it after the Fire and Mello paper [in Nature, and] it was obviously fascinating at that point. In the first year after that there were still only, at most, about 10 papers in the field — any field that only has about 10 papers a year is a field that I feel I can follow.
It was just something I was following out of curiosity for a long time, and we’d been working on asymmetric cell division in C. elegans. I had a graduate student who was starting, had experience in C. elegans before, and who was interested in doing an RNAi-based screen for genes involved in asymmetric division. This is Nate Dudley.
He started a screen where we did a trick that we’d learned from Bruce Bowerman’s lab [at the University of Oregon]. Bruce Bowerman’s lab had done screens where they’d injected eight double-stranded RNAs at a time [into an organism] and looked for interesting phenotypes. So we were going to do this and look for genes required for asymmetric cell division.
What happened was, Nate started injecting pools of eight double-stranded RNAs, and on his second pool he was surprised that a known embryonic lethal didn’t come up as lethal. What happened was, in this initial experiment there were 10 pools of eight double-stranded RNAs each and we left known embryonic lethals in there to make sure that we could recover them so that essential genes are giving what we expect — this is sort of a positive control to know that the double-stranded RNAs are working as they should.
In pool number two, he got no lethality and yet we knew there was a double-stranded RNA for an essential gene in there. Nate came to me and said: ‘I got no lethality for pool two.’
When he first came and told me this, I didn’t realize one of the lethals was in there. So the first thing I said to him was: ‘Why don’t you move on to pool number three,’ which seemed obvious at the time. He said: ‘Wait a second, there’s a known lethal in there and it’s not coming up as lethal.’ I said that probably the double-stranded RNAs aren’t made as well as we’d hoped, maybe some of them aren’t working. There had been rumors [also] that you can get non-specific suppression when you inject too many double-stranded RNAs at a time.
He was a couple steps ahead of me and had actually taken that pool and just injected the RNA corresponding to the essential gene, and that gave near 100 percent embryonic lethality. So, on its own it was working, but in the group it wasn’t.
The essential gene was called GLP-1 and Nate said: ‘I think I have a suppressor of GLP-1 RNA.’ I said: ‘That’s crazy.’ You wouldn’t expect among a group of eight genes — genes chosen almost at random — to find a suppressor of GLP-1. I said that probably something else was going on.
He was right, in a way: It was a suppressor of GLP-1 RNAi, but it wasn’t a suppressor of GLP-1; it was a suppressor of RNAi.
The way we figured that out was that first he figured out which other double-stranded RNA was suppressing this effect, and it turned out that there was one that had a potent suppressing effect and none of the others did. He tried injecting that into GLP-1 mutants and it had no effect at all; it wasn’t rescuing the mutants, it was only suppressing the RNAi effect.
Then, to test whether it really suppresses RNAi, we co-injected that double-stranded RNA with double-stranded RNAs corresponding to a bunch of essential genes and found that they were all suppressed, although some to a greater extent than others. What we had then was that we were looking for genes required for asymmetric cell division and what we’d found is a potent suppressor of RNAi.
At that point, we thought that this could be a really easy way to identify genes for RNAi and that’s where we started to change paths. In fact, he did one more experiment, which was to try the same experiment with known components of the RNAi mechanism. As [Dudley] was doing this, Greg Hannon’s lab [at Cold Spring Harbor Laboratory] showed [that] co-transfecting Dicer double-stranded RNA with another double-stranded RNA into fly S2 cells can weaken the RNAi effect.
Nate still had an interest in asymmetric division, and I worked pretty hard to convince him that he should be studying RNAi from now on.
Where has this all led to in your lab?
What happened was that he’d identified this gene and we were curious how it worked. The great thing about identifying genes using RNAi is that you know the genes you’re working with right from the start — any time you identify something interesting, you go right to your table instead of spending months trying to clone the gene that a mutation found.
When we looked at the sequence of this thing, we were scratching our heads. The sequence of it [made it] look like a human gene called GAS41 — this is a gene that had been implicated in glioblastomas.
I should say that the gene we identified we then called GFL-1. It has a double meaning: It’s for GAS41-like, but gaffle is urban slang for ‘trick’ or ‘steal’, and it was the double-stranded RNA that was tricking us.
GFL-1 looks like the human gene GAS41, and they both have domains that suggest they interact with chromatin. GAS41 in humans actually has a binding partner for which there’s a really good C. elegans homolog. Nate did the same kinds of experiments with the C. elegans homolog — the co-injection of double-stranded RNAs — and found that he got the same effect with that one, too.
So, we had two C. elegans genes that we thought might be involved in RNAi because when you do RNAi to them you turn off the RNAi mechanism.
I should say that when we started thinking about this, at first it didn’t make sense: that you could use RNAi to turn off RNAi, that you could use a mechanism to turn off a mechanism because right when you need it, it’s not present. But with a little more thinking we realized that when you inject a double-stranded RNA corresponding to an essential part of the RNAi machinery, you’re not immediately getting rid of the protein, you’re getting rid of the RNA and then the protein will disappear at some rate dependent on how fast it’s normally made and normally degraded in cells. If you’re co-injecting a double-stranded RNA corresponding to an essential gene, that protein also has a rate at which it’s degraded. As long as the balance between those two is right, you’d expect this works. We basically expected that some genes would make very good double-stranded RNAs to co-inject [in order] to identify these things and some wouldn’t.
What we did was find ones that worked well for the first gene, and those have worked consistently for others.
So, we had these two genes and we were still confused because both of these were expected to be nuclear proteins and, at the time, RNAi was a cytoplasmic phenomenon — everything we knew about how it worked, at least in animal cells, suggested that it was working in the cytoplasm degrading mRNAs.
We looked at the sequence of these two things and realized that one of them, GAS41, had a sequence that looked like something that could interact with polycomb proteins. Now, C. elegans doesn’t have very many polycomb group proteins — it’s missing a big chunk of the set that’s present in humans and flies. But it does have some, and Susan Strome’s lab [at Indiana University] had been working on these for a while.
We tried co-injecting double-stranded RNAs for the polycomb group genes with [the] double-stranded RNAs corresponding to essential genes, and those gave us the same effect, at least for some of the polycomb group proteins — for three of them, but not a fourth.
At this point, all of our experiments were implicating these genes in functioning in RNAi by co-injecting double-stranded RNAs, and we felt it was important to nail [down] that a mutant that’s lacking one of these genes really doesn’t do RNAi. For the polycomb group this is MES-6, and a couple of others that have been implicated in the same pathway: MES-4 and MES-3.
Co-injection … suggested that all of those [genes] are required for RNAi. When we injected the double-stranded RNA into the mutants, those also failed in RNAi. Now, we’ve got up to five nuclear proteins required for RNAi, which didn’t confuse us any less.
The one bell that this rung in our heads is that in plants it had been shown that double-stranded RNAs can function to silence genes in the nucleus. So we that that maybe what’s happening is that when you introduce double-stranded RNAs into animal cells, as well as eliminating the pool of double-stranded RNA, you also turn off the faucet, you turn off the transcription of targeted genes in the nucleus.
In yeast, this was shown recently by a few labs to be the case. But in animal cells, it’s not completely clear if this is the case or not. The closest thing we have to whether or not it’s clear is the [Jim] Birchler paper [in Science], which shows that normal silencing in position effect variegation depends on RNAi components. (See RNAi News, 1/30/2004).
But I think it’s still not completely clear whether when you introduce a double-stranded RNA, does the locus your targeting get silenced in animal cells. Of course, animal cells are what we’re really interested in because we’re animals — because of therapeutic possibilities.
So you’re just continuing that effort now?
There was some strange phenomenon that came up with these MES genes that we’re curious about. One phenomenon was that if you inject a typical concentration of double-stranded RNA, the MES mutants — these polycomb mutants — have trouble performing RNAi. If you inject very, very little RNA, at least in some of these mutants, you get a stronger effect. This suggests to us that something unique happens when you inject very little double-stranded RNA.
So, we’re wondering if amplification mechanisms that other labs have discovered kick in only in the presence of low amounts of double-stranded RNA, and the amplification mechanisms can somehow obviate the need for nuclear silencing.
That’s one area of interest, but most our effort now is in another area, which is using the same screen to identify other genes involved in RNAi. We think we have a really efficient mechanism for identifying genes required for RNAi in C. elegans.
Can you describe that screen in any sort of detail?
We realized that it would be reasonable to go through the genome and pick out our favorite 50 to 100 candidates, based on sequence, for genes that might be involved in RNAi, and then to do co-injection experiments.
So Nate’s doing experiments where he’ll do these co-injections for genes of interest, and then we go through the same sort of process — once we’ve got a positive that way, we wait until we have a mutant to make a conclusion about whether the gene’s involved in RNAi or not.
So he’s collecting a list of genes we think are implicated in RNAi, and he’s starting to work on how certain of these are functioning in RNAi.