More Recent Comments

Friday, January 14, 2011

A Challenge to Fans of Alternative Splicing

There are many well-known examples of alternative splicing. These examples have been taught in undergraduate courses for 30 years and they are prominently featured in textbooks. Alternative splicing exists.

Here's the problem. The explosion of EST data in the 1990's resulted in detection of many sequences suggesting that alternative splicing was much more widespread that previously suspected. The vast majority of these claims have not been verified and many of them have been removed from the annotated genomes published in the past few years.

Now we have a whole new set of claims based on high throughput analysis of transcripts from a variety of organisms and tissues. Many workers believe that the majority of human genes are alternatively spliced and some even publish articles stating that 95% of humans genes exhibit alternative splicing. One of my colleagues who makes such a claim says that just because a gene is alternatively spliced doesn't mean that the various isoforms of RNA are functional but I think that's disingenuous. If it's going to be a meaningful term then "alternative splicing" has to imply that that at least two different versions of RNA have some biological function.

I've asked repeatedly for evidence that some particular genes are alternatively spliced to give rise to two or more functional products. It should be possible to get this information from the databases used by these researchers—the ones that support their claim of widespread alternative splicing. Unfortunately, this has proven to be difficult. Whenever I search common alternative splicing databases I'm told that those databases aren't very good and the results aren't reliable.

Here's the challenge to all researchers who believe that a majority of human genes are alternatively spliced (in a biologically relevant manner). Show me your data. Pick one of the following sets of genes and demonstrate that most of them have functional alternatively spliced transcripts. If none of the genes in the set qualifies then explain why you reject the presumed alternative transcripts shown in popular databases. This shouldn't be much of a challenge if your claim is correct.

Note that this is a two part challenge. You have to first present evidence that there are functional alternative splicing events and then you also have to present the reasons why you reject some of the data from sequenced RNAs.

Here's an example from the human gene for triose phosphate isomerase (TPI1). The Entrez Gene entry is Gene ID: 7167. The primary entry shows one alternatively spliced transcript that removes the N-terminal coding region of the protein and creates an new larger N-terminal sequence. What is the evidence that this is biologically relevant? Now check out the known transcripts that have been detected according to UCSC Genome Browser, AceView, and Model Maker. These show additional variants affecting the splice junction sequences around exons 2, 4, and 6. Are these also examples of alternative splicing? Are they functional? If you reject these variants then what's the rationale for accepting some possible transcripts as real but rejecting others? Are some of them artifacts?

The three gene sets are ...
  1. Human genes for the enzymes of glycolysis.
  2. Human genes for the subunits of RNA polymerase with an emphasis on the large conserved subunits [Two Examples of Alternative Splicing]
  3. Human genes for ribosomal proteins.

I selected these examples because we know the structures of the proteins so we can evaluate the possibility that an alternatively spliced message might produce a novel polypeptide chain. Of course there might be other reasons (regulation?) for producing alternatively spliced transcripts. Feel free to present the evidence.

Now it's possible that I've accidentally chosen sets of genes that do not exhibit alternative splicing. If that's the case then pick any other set of genes with a common function where the structure of the protein product is known. Meanwhile, you can explain why you reject all the putative splice variants for these genes.


26 comments :

DK said...

Better yet, any of the many actin genes. All of them contain 5-10 exons and are claimed to be alternatively splicied. But actin structure is essentially untouchable. There is no way any of these transcripts are functional.

OTOH, tropomyosins represent a nice clear-cut case of the importance and functional significance of alternative splicing.

Anonymous said...

The word "functional" requires a human judgement call. For example, does "keeping the nonsense-mediated decay system busy" count as a function?

It might be more precise to require the ASTs be under positive selection.

Larry Moran said...

anonymous says,

The word "functional" requires a human judgment call. For example, does "keeping the nonsense-mediated decay system busy" count as a function?

"'When I use a word,' Humpty Dumpty said, in rather a scornful tone, 'it means just what I choose it to mean — neither more nor less.'

'The question is,' said Alice, 'whether you can make words mean so many different things.'

'The question is,' said Humpty Dumpty, 'which is to be master — that's all.'"

Keeping a repair system busy might be an important "function" in Wonderland, but not in a rational world.

Anonymous said...

Keeping a repair system busy might be an important "function" in Wonderland, but not in a rational world.

And... you know this how? Or is this just your opinion? What is your cut-off for deciding on the "importance" of a function? If you're going to throw down the gauntlet, it's only fair that you objectively define your criteria for the challenge.

Arthur Hunt said...

The use of alternative splicing to fine-tune overall RNA expression levels is really no different in principle from the use of the ubiquitin system to fine-tune protein levels. In this respect, "non-productive" splicing most certainly can be functional.

I think a fruitful experimental approach to this question is to replace, by any of a number of means, an endogenous gene with ones that cannot be subject to specific sorts of alternative splicing and see what the consequences are. It's pretty easy to scour the literature and find cases that clearly show that alternative splicing is needed for the functioning of a given gene.

But I am curious - Larry, how many experiments such as I outline are you aware of where the result is negative, where alternative splicing is not implicated in the functioning of a given gene? There must be many, many hundreds, given your skepticism. Perhaps you could give us a few examples.

Larry Moran said...

Arthur Hunt says,

The use of alternative splicing to fine-tune overall RNA expression levels is really no different in principle from the use of the ubiquitin system to fine-tune protein levels. In this respect, "non-productive" splicing most certainly can be functional.

The ubiquitin system has a function and so does nonsense mediated decay. Their function is to eliminate aberrant products. However, it doesn't make any sense to say that the mistakes giving rise to those non-functional products is also a function.

Do you really want to claim that splicing errors are "functional" because they keep the repair system busy?

I think a fruitful experimental approach to this question is to replace, by any of a number of means, an endogenous gene with ones that cannot be subject to specific sorts of alternative splicing and see what the consequences are. It's pretty easy to scour the literature and find cases that clearly show that alternative splicing is needed for the functioning of a given gene.

I agree. There are quite a few excellent examples. Do you think there's evidence that a majority of human genes produce functional alternatively spliced products? If the experiments haven't been done then why are people making such claims?

But I am curious - Larry, how many experiments such as I outline are you aware of where the result is negative, where alternative splicing is not implicated in the functioning of a given gene? There must be many, many hundreds, given your skepticism. Perhaps you could give us a few examples.

You know very well that: (a) you can't ever prove a negative, and (b) experiments showing that something doesn't happen don't get published.

But you're missing the important point. The burden of proof is on those making the claim that most human genes exhibit alternative splicing. I'm skeptical of that claim but that doesn't mean I have to prove the negative.

That's not how science works.

Larry Moran said...

anonymous spouts,

And... you know this how? Or is this just your opinion? What is your cut-off for deciding on the "importance" of a function? If you're going to throw down the gauntlet, it's only fair that you objectively define your criteria for the challenge.

We're talking about common sense definitions of words. What you're saying is equivalent to the claim that car accidents have an important function—it's to keep hospitals busy.

Larry Moran said...

Has anyone noticed that there are some fans of alternative splicing who want to quibble about definitions but nobody has even attempted to meet the challenge?

I wonder why? Is it that difficult?

DK said...

Has anyone noticed

Some certainly have. One of the truly weird things about it is the refusal to think logically:

- It is known that noise exists; from all we know, it must exist too.
- It is known that the methods to detect many variants are not very quantitative and don't distinguish well between signal and noise.
- Therefore all the detected signal must be non-noise.

WTF?

Arthur Hunt said...

The ubiquitin system has a function and so does nonsense mediated decay. Their function is to eliminate aberrant products. However, it doesn't make any sense to say that the mistakes giving rise to those non-functional products is also a function.

Do you really want to claim that splicing errors are "functional" because they keep the repair system busy?


Um, these "repair systems" do much more than just remove damaged goods. It's pretty well-established that post-transcriptional and post-translational negative controls function in essential ways to tune overall levels of gene expression. And in ways that have nothing to do with quality control. This goes for NMD (and ubiquitin).

You know very well that: (a) you can't ever prove a negative, and (b) experiments showing that something doesn't happen don't get published.

The reason I asked about "the experiments" (that often do support roles for alternative splicing) is that you have claimed, elsewhere, that probably as little few as 5% of mammalian genes actually are alternatively-spliced in functionally-significant ways. I would have thought, from claims like yours, Larry, that you could offer up 20 counter-examples for every one that others can point to that shows a function for alternative splicing. Your answer to my question tells me that the 5% number is just something you made up, without regard to evidence. In other words, purely opinion and speculation.

That's OK, there's nothing wrong with opinion.

Oh, and negative results do get published. And, in this field, negative results of experiments such as I described would indeed be publishable.

Has anyone noticed that there are some fans of alternative splicing who want to quibble about definitions but nobody has even attempted to meet the challenge?

The "challenge" is, upon reading, pretty garbled:

Note that this is a two part challenge. You have to first present evidence that there are functional alternative splicing events and then you also have to present the reasons why you reject some of the data from sequenced RNAs.

The first part has been met, and you agree (apparently) that good evidence exists for a function for alternative splicing in gene expression. The second part is diametrically opposed to the first part, and as presented implies that the "data from sequenced RNAs" in some way argues against a function for alternative splicing. This makes no sense (at least to me).

Arthur Hunt said...

Has anyone noticed that there are some fans of alternative splicing who want to quibble about definitions but nobody has even attempted to meet the challenge?

The "challenge" is, upon reading, pretty garbled:

Note that this is a two part challenge. You have to first present evidence that there are functional alternative splicing events and then you also have to present the reasons why you reject some of the data from sequenced RNAs.

The first part has been met, and you agree (apparently) that good evidence exists for a function for alternative splicing in gene expression. The second part is diametrically opposed to the first part, and as presented implies that the "data from sequenced RNAs" in some way argues against a function for alternative splicing. This makes no sense (at least to me).

Anonymous said...

Here's a functional role for the transcripts of pseudogenes anyway, if articles in Nature are to believed:

"A coding-independent function of gene
and pseudogene mRNAs regulates
tumour biology"
http://www.nature.com/nature/journal/v465/n7301/pdf/nature09144.pdf
(from 24-Jun-2010)

It's not much a stretch to infer similar function for alternative transcripts of known functional genes.

Arthur Hunt said...

Sorry about the partial double post. The system asked me to trim my first comment, which I did. But the system kept and submitted the comment that was rejected.

There may be another comment of mine with essentially identical thoughts lost in limbo - if it pops up, ignore it.

Again, apologies.

Larry Moran said...

anonymous says,

Here's a functional role for the transcripts of pseudogenes anyway, if articles in Nature are to believed:

Gould and Lewontin pointed out in 1979 that most arguments are not over the existence of a given phenomenon but over its relative frequency and importance. Some people are confused about this.

Nobody is surprised to find examples of pseudogenes that have acquired some useful function. That's not the point. The point is whether you can justify the existence of pseudogenes by claiming that they might, at some time in the future, acquire a new function.

Is it a general argument?

It's not much a stretch to infer similar function for alternative transcripts of known functional genes.

It's a huge stretch to assume that most genes exhibit functional alternative splicing just because some do.

Larry Moran said...

Arthur Hunt says,

The first part [of the challenge] has been met, and you agree (apparently) that good evidence exists for a function for alternative splicing in gene expression.

Here's the first part of the challenge.

Pick one of the following sets of genes and demonstrate that most of them have functional alternatively spliced transcripts.

Did you post an answer that got lost? If so, please re-post it.

Larry Moran said...

Arthur Hunt says,

It's pretty well-established that post-transcriptional and post-translational negative controls function in essential ways to tune overall levels of gene expression.

I agree. What has this got to do with alternative splicing? Are you arguing that the majority of genes produce 5 different transcripts in order to fine tune gene expression? How does that work using one of the sets of genes that form part of the challenge?

Your answer to my question tells me that the 5% number is just something you made up, without regard to evidence. In other words, purely opinion and speculation.

That's pretty close to being correct. This is why I issued the challenge. I'm looking for someone who will step up to the plate and demonstrate that a majority of a given set of genes exhibits functional alternative splicing.

I'm basing my 5% estimate on known examples of alternative splicing. I doubt very much whether there are really 1000 known examples but I decided to be generous.

How many proven examples are there according to your reading of the data?

Physeter said...

Anon:
The word "functional" requires a human judgement call.

Arguably yes, but does any human actually defend that "keeping the nonsense-mediated decay system busy" counts as a function?

Larry:
Keeping a repair system busy might be an important "function" in Wonderland, but not in a rational world.

It is not about being important, but being a function at all, isn't it?

Physeter said...

What you're saying is equivalent to the claim that car accidents have an important function—it's to keep hospitals busy.

Reminds me of this movie.

Anonymous said...

There was an article where the authors sought for possible alternative splicing sites in ESTs (if memory does not fail), and built a database on domain boundaries (based on structures and probably protein families). In the end they found that alternative splicing occurred twice as many times as expected by chance between such boundaries. They concluded that alternative splicing's was there to produce functional variants. The little tiny detail was: the total number of alternative splicings falling between domains was a little bit above 4% (baseline was a bit above 2%). I would not conclude that alternative splicing is there for providing proteins with different functions from such poor data.

Anyway, I am with you Larry. I accept that there are clear cases where this thing is important, but I don't buy on it being as abundant as people might want it to be. I also think this issue about alternative splicing and protein variants is an attempt at consolation that we have "few" genes.

The Other Jim said...

One fun experiment is to compare the alt-splicing predictions between a given mouse and human homologue. Typically, it's the human one (usually based on an assay in a tumour cell line) that has all of alt-spice predictions, and the mouse one (from a dissected tissue) has very few, if any.

Hmmmm. Co-incidence?

Larry Moran said...

The other Jim says,

One fun experiment is to compare the alt-splicing predictions between a given mouse and human homologue. Typically, it's the human one (usually based on an assay in a tumour cell line) that has all of alt-spice predictions, and the mouse one (from a dissected tissue) has very few, if any.

Hmmmm. Co-incidence?


You are obviously not a true believer. If you were, you would recognized that this observation explains why humans are far more complex than mice in spite of the fact we have about the same number of genes.

Humans have evolved alternative splicing in order to make a whole bunch of extra proteins from the same gene. These extra proteins explain why we are smarter than mice! :-)

Anonymous said...

Larry,

I think you are mixing two different things - the fraction of alternatively spliced transcripts that are functional, versus the fraction of human genes that contain functional alternatively spliced isoforms. While the majority of transcript isoforms detected by EST or RNA-Seq could be non-functional forms or splicing errors, the percentage of human genes with known functional (or regulated) alternative splicing events is substantial and steadily increasing. For example, ~90% of multi-exon genes express minor isoforms that account for >15% of the total transcripts, so this cannot simply be attributed to rare spliceosome errors. There are also well-documented examples of global shifts in alternative splicing regulation during key biological processes such as neuronal development (Genes Dev. 2007 Jul 1;21(13):1636-52) and epithelial-mesenchymal transition (EMBO J. 2010 Oct 6;29(19):3286-300). In fact, any single splicing regulator is typically responsible for regulated alternative splicing events in hundreds to thousands of genes.

Larry Moran said...

anonymous says,

For example, ~90% of multi-exon genes express minor isoforms that account for >15% of the total transcripts, so this cannot simply be attributed to rare spliceosome errors.

I don't believe that but your argument would be much more convincing if you rose to the challenge I presented.

Why not pick one of the categories and explain why you think the alternatively spliced transcripts are functionally important? How hard can it be to meet the challenge?

Why is nobody even trying?

Gireesh K.Bogu said...

Because the people who can answer this are busy with their work :)

However it is very simple challenge if you use RNA-Seq data and MISO (http://genes.mit.edu/burgelab/miso/) by comparing with one RNA-Seq with another to filter alternative splicing events.

For example if your interested gene is expressed in human brain then take human brain RNA-Seq data and compare with other tissues of human (For this you can use ERIC.T.WANG data, Nature).

I'm sorry that I'm just posting simple solution for the question instead of doing it. As I said I'm busy with my work.

Hope I addressed your question

Larry Moran said...

Gireesh K.Bogu says,

I'm sorry that I'm just posting simple solution for the question instead of doing it. As I said I'm busy with my work.

That doesn't even come close to meeting the challenge.

If you are a believer in abundant alternative splicing, and you work in the field, then what kind of "work" could possibly be more important than checking to see if your hypothesis makes any sense?

I'm sick and tired of hearing excuses from the labs that are promoting this nonsense. What are you afraid of?

Anonymous said...

Hi Larry,

Although I generally agree with your post since I found it difficult to find examples, the most known example that I am aware of is sex determination in flies http://nar.oxfordjournals.org/content/40/1/1.long

Looking forward to reading your response,
Thanks