The cost of garbage
January 10, 2011 § 5 Comments
Misfolded proteins are bad news. Not only are they involved in a number of nasty diseases, they also place potentially severe constraints on evolution. As we’ve discussed before, evolution depends on the ability to survive and function in the face of mutations. A mutation that causes misfolding, so that instead of a nice functional protein you get gunk, causes a number of important problems. First, you’ve wasted all your effort in transcribing an mRNA from a gene, and then translating the mRNA to produce a protein. Second, you still have to make another one. Third, you have to get rid of the gunk, otherwise it may clog up essential functions (as uncollected trash does in 
New York City). Even if the increase in gunk production as a result of your mutation is modest — suppose the mutation leads to misfolding of one in every 100 protein molecules, for example — over the millennia the difference between a cell that has to deal with a lot of misfolding and a cell that has less of a problem could add up. And if a mutation has a beneficial effect, but also causes an increase in misfolding, the beneficial effect would have to be strong enough to overcome the negative effects of misfolding for the mutation to be positively selected.
How large a problem is a misfolded protein for the cell, and what matters more, the diversion of protein production capacity or the need to get rid of the gunk? A recent paper from Allan Drummond’s lab (Geiler-Samerotte et al. 2010, Misfolded proteins impose a dosage-dependent fitness cost and trigger a cytosolic unfolded protein response in yeast, PNAS doi/10.1073/pnas.1017570108) reports the results of a determined and careful effort to find out.
Working in budding yeast, Geiler-Samerotte et al. used two paired sets of wild-type proteins and misfolding variants for their studies. The first set is based on YFP, a yellow version of green fluorescent protein, while the second is based on orotidine-5′-phosphate decarboxylase (URA3), a protein in the uracil synthesis pathway. (They add wild-type YFP to the URA3 in order to be able to track cells carrying it.) Each of the wild-type or misfolding variant proteins (four for YFP, one for URA3) is placed under a promoter whose expression level can be tuned by adding various levels of galactose. YFP has no function in budding yeast, and URA3 can be made unnecessary by simply adding uracil to the medium. This is important, because it allows the authors to ignore the fitness effect of not having a functional protein, and focus instead on the cost of making the protein and the cost of gunk removal.
First question: does expressing a misfolded protein indeed have a fitness cost? Fitness is something that can only be measured in competition, and the fitness costs that could be important in evolution are very very small. To measure fitness costs with high accuracy, Geiler-Samerotte et al. competed each of their YFP-expressing strains (wild-type and mutants, all of which are yellow) against an RFP-expressing strain (red) for 17 generations, counting the number of cells that are yellow and the number of cells that are red at each generation using FACS. The strains expressing the four different misfolding mutants they used all grew significantly less well than the control strain carrying wild-type YFP, showing a growth disadvantage ranging from 0.7% to 3.2%. Similarly, the destabilized URA3 mutant had a growth disadvantage relative to wild-type of about 1.2%. To understand why the different mutants had different effects, Geiler-Samerotte et al. measured the amount of insoluble protein in each YFP strain: plotting growth disadvantage versus insoluble protein gave an almost linear relationship. So the mutants that most effectively destabilize protein structure are also the ones that make the most difference to growth rate.
How large is the effect of expressing a useless protein? The authors can measure this by comparing the growth of strains that are induced to express the wild-type (but useless) protein with the growth of uninduced strains; this is done by varying the amount of inducing galactose in the media. For both wild-type YFP and wild-type URA3-YFP, the growth disadvantage resulting from expression is 1.4%. This number is interesting because the authors also measured what percentage of the cellular protein the YFP represents, and it’s small: 0.1%. Can diversion of 0.1% of protein synthesis capacity really have a 1.4% effect on growth rate? The authors point out, though, that this is the maximum possible growth disadvantage of expression; if the wild-type protein is misfolded to some degree, this number would be a combination of gene expression cost plus the misfolding cost. Allan tells me that there’s reason to believe that even the wild-type YFP isn’t fully folded.
Returning to misfolding, let me just emphasize the numbers here. The growth disadvantage of the extra misfolding caused by the mutants, on top of the expression cost, is 0.7% to 3.2%, as I said before. This is the cost of misfolding some fraction of a protein that is only expressed at 0.1% of total cellular protein. Wow. That’s pretty nasty gunk. One would assume that something that poisonous would generate some kind of response from the cell, and it does. The authors used quantitative mass spectrometry to measure the levels of over 3000 proteins (half the yeast proteome), and asked what changes when a misfolded protein is expressed. They found 25 proteins that showed significant changes, almost all of which appear to be part of the heat-shock response. This makes total sense — most of the upregulated proteins are chaperones that would be expected to catalyze the rescue and refolding of insoluble misfolded proteins. (Not every chaperone you might expect to be mobilized shows up, though; this system might offer a useful window on selective control of chaperone expression.)
You could argue that it’s possible that the gunk itself isn’t harmful, and that the cost comes solely from expressing all those chaperones — the chaperones evolved, let’s say, for another purpose (e.g. to deal with heat shock) and the misfolded proteins are setting off the response by accident. [This does happen in biology. Your immune system didn’t evolve to respond to pollen, and there’s no functional reason for it to do so, but — as you know if you suffer from allergies — the misdirected response can be quite powerful.] The other possibility is that the cost of producing chaperones is, for the cell, a lesser evil than leaving the misfolded protein lying around. You could test this by suppressing chaperone induction and asking what happens to the fitness of cells expressing misfolded mutants. If the misfolded protein is toxic by itself, the cells should be sicker than when you allow the chaperone response to deal with them. Geiler-Samerotte et al. haven’t done this experiment (or at least, it’s not reported in this paper; I’d bet it’s in progress), but they make a strong prediction that blocking chaperone induction will increase, not decrease, the fitness cost of misfolding.
Either way, it’s clear that the fitness costs of mutations that destabilize protein structure are significant, and easily large enough to make a difference over evolutionary time. Getting out my compound interest calculator, I see that a 3% difference in growth per generation will translate into a ~20-fold larger population within 100 generations. For yeast, that works out at a couple of weeks of exponential growth. I don’t know about you, but this paper leaves me convinced that avoiding misfolding is likely to be a surprisingly strong selective pressure in evolution.
Geiler-Samerotte KA, Dion MF, Budnik BA, Wang SM, Hartl DL, & Drummond DA (2010). Misfolded proteins impose a dosage-dependent fitness cost and trigger a cytosolic unfolded protein response in yeast. Proceedings of the National Academy of Sciences of the United States of America PMID: 21187411
It Takes 30 
Great stuff!
I wonder if it relates somehow to Uri Alon’s piece on the cost of the unneeded proteins, where he shows in bacterial system that the price depends on the growth phase (I discussed it in my blog just a day ago: http://stringentresponse.blogspot.com/2011/01/you-pay-in-ribosomes-for-proteins.html ). They use stable GFP in their approach, and see that price is much higher in the early exponential phase, and goes dramatically down later.
So the question is – is GFP equally stable in these two conditions (and ergo has the same price per molecule)? If not, the effect they see could be similar in nature to what is described in this piece. Nature of the bacterial cytoplasm does change under different conditions (redox potential, pH) – this in turn could affect the stability, I would guess.
Vasili — Your question is dead-on, and your observation about cytoplasmic changes possibly influencing folding is really interesting.
It’s worth pointing out that the experiments are very different: our experiments were done in yeast expressing a protein from a single chromosomal locus, whereas the Shachrai et al. experiments were done in E. coli expressing a protein from a high-copy plasmid (typically tens of copies). They report production up to 15% of total protein; this would make their GFPmut3 the top-ranked protein by abundance in the cell. We report YFP production of less than 0.1% of total protein, and YFP is about 150th in abundance. These production regimes are different enough that the dominant biological costs may well be distinct.
Still, one wonders if the E.coli cells are up-regulating chaperones as part of the cost-mitigating program. We know GroEL comes up in response to overexpression of lacZ by Vitkup’s group (http://genomebiology.com/2010/11/9/R98) and earlier by Kurland’s group. Shachrai et al. suggest this: “Another alternative is that the limiting factor in the first phase is not ribosomes but other factors, such as chaperones or other cellular machinery, perhaps also regulated by ppGpp.” Plenty of work’s being done on these questions, so we won’t have to wait long, I’ll wager.
Allan – in Plata at al. (the paper from Vitkup’s group you refer to) they claim that misfolding in E. coli does not contribute significantly to cost of gratuitous protein expression. This seemingly dismisses my concern about GFP misfolding effect contributing to changes in protein expression const during early and late exponential phase observed in Shachrai et al.
Or am I missing something here? I would guess that assays used in Plata at al. and in your paper differ radically in the sense that you have much better estimate of the protein expression level then they do, which is kind of important for comparing the cost normalized per unit of protein produced.
We address the Plata et al. study in our paper (first full paragraph on p.5); in essence, we observe that overexpression of the target construct in that work, even of the folded WT enzyme, pushes cells into a stress response, and that it’s unclear whether the fitness effects one observes during that response tell you much about what the fitness effects would be in the absence of such a response (as in our work). In the Shachrai et al. study, if GFP acts like lacZ, then yes, I’d expect that your concern has been addressed — for the expression regime where GFP is 15% of total protein. However, I think your concerns about GFP misfolding remain interesting and should be tested in a lower expression regime, the regime in which most proteins are expressed under normal conditions.
Yep, i’ve read this part in your paper, but I thought there might be more to it (I always have better explanations when I am reading the proofs).
Would be of interest to see what you get with GFP.