iGEM2011: GFP-based readout for ppGpp concentration in vivo

Measuring ppGpp concentration in individual living cells with good temporal resolution would be great. I've been musing about a possibility of doing that using RNA aptamers, but that's just musing. It seems like am iGEM team from the University of Trondheim tried setting up a GFP-based reported system, and this system is, maybe, possibly, probably, working. Somewhat. 


Unfortunately there is no publication. However, there is a popular article about the whole affair, in Norwegian and there is a short report on the iGEM webpage. 


Apropos to the technical issues that are listed in the original report, such as dramatic leakage of the GFP reporter in the absence of stress stimuli, there are several conceptual concerns. First, the system is based on translation of the GFP reporter during the stringent response, and during the stringent response translation is strongly inhibited. Second, GFP (they use red version of it, mCherry) has to mature in order to become bringt, and that takes some time - from minutes to hours, depending on the conditions and what sort of GFP variant it is. For mCherry maturation time is 15-40 minutes, and this is comparable with E. coli generation time. Therefore, first, one would expect a very pronounced lag before the SR is engaged and the corresponding readout and, second, all the fluctuations in the ppGpp concentrations happening on the timescale below 10s of minutes will be averaged out. Third, GFP is very stable, so this reporter system will have severe memory effects - once the cell has committed to stringency, it will produce GFP, and even though stringency is reversed, GFP will stay. Maybe it is possible to turn this into a feature, but I am not sure how. And, lastly, brightness of the GFP depends on the pH and redox potential of the environment, and these things change in the stressed cells.

photoconvertable GFP variants for SPT

Many different fluorescent labels are used for super-resolution imaging, and photoconvertable GTP variants are probably the most widely used. And again, there are many photoconvertable GFP variants: PATagRFP, PA-GFP, Dendra, EosFP and many others.

What makes a good photoconvertable GFP for single particle tracking (SPT)? Well, all the things that make a good label (bright, stable, fast to mature, small, inert, this convenient and nicely separated adsorbance and emission spectra - the obvious stuff) and one more thing that is not so obvious.

Photoconvertable label should not convert by itself. It should have a stable dark state and conversion should be strictly upon illumination by the photo-converting light. Otherwise creating single molecules for SPT is hard - they just pop into existence by themselves! In than sense EosFP is much better than Dendra.

Sounds obvious, but it wasn't to me...

You pay in ribosomes for proteins

Everything costs. When cell grows, it needs energy and in needs materials. By the end of the day it comes down to accounting: if you need to make N proteins, you will need X ATPs molecules, Y aminoacids and Z ribosomes to do the job. And of all these ribosomes are the most expensive to make: they are huge, made of RNA and if you want to make proteins fast, you need lots of ribosomes!

So research team led by famous systems biologist Uri Alon decided to quantify the cost of making a protein. In order to do so, they forced E. coli producing Green Fluorescent Protein (GFP) which is inert and easy to quantify. They figured out that early in the exponential phase the cost of GFP (that is decrease in growth rate associated with production of a given amount of GFP) is high, but later on it markedly recreases!

This observation makes immediate sense: first you need to produce the tools for producing GFP (ribosomes, energy in the form of ATP etc.) and if instead of this you make GFP this GFP comes at a high prise. Corroborating with this logic, they figured out that if you transfer bacteria from energy-reach media to energy-poor, the price for GFP is low: well, you accumulated all these ribosomes during the good times, so now you can make some GFP cheap.

There could be an interesting connection here with another resent paper where in yeast it was shown that the cost of GFP is dramatically different for stable and denaturation-prone variants (for brilliant discussion of this paper see this post in It Takes 30). Is GFP equally stable in E. coli during the early and late exponential phase? Could it be that the effects observed here are reflecting mere change in GFP stability? Intracellular conditions do change in E. coli under different conditions, so it is possible that GFP is not always equally stable, and this may affect its physiological cost. Surprisingly, another report claims that in E. coli aggregated and soluble LacZ have very similar cost, which to some extent dispels my worries about GFP stability and cost.

The read-out for GFP quantification could be affected by cellular milieu as well: judging from my experience, GFP is definitely not always equally bright, and this could affect estimates of GFP concentration and thus the estimates of its physiological cost.

Discovering differences in the GFP cost in early and late exponential phase prompted the authors to try figuring out what cellular system is behind it. And they had a very good initial guess. In bacteria adaptation to changes in availability of food are regulated by the stringent response mechanisms, with RelA and SpoT proteins doing the job (see my previous posts on that subject, see 1 and 2). RelA produces ppGpp molecule that compels the cell to stop producing ribosomes and concentrate on aminoacid production, and SpoT mostly degrades ppGpp: too much of it would lead to complete inhibition of ribosomal production and eventually - death.

Therefore it is only logical that in the paper in question different knock-out strains missing RelA and SpoT were tested, and indeed, stringent response machinery turned out to the the key to radical changes of the protein cost during different stages of bacterial growth.

And here is the catch.

One of the strains they used was SpoT(-) RelA(+) strain, that is one having NO SpoT and allegedly INTACT RelA. As we discussed above this bug should be very much dead, and as yet no one managed to produce this strain. So what's about the strain presented in the paper then?

Well, there are many options. When you want, really want to knock out a gene, you finally succeed. However, bacteria want to live, and you select the ones with mutations that compensate for the knock-out of the gene you have. For instance, you can mutate main target of ppGpp, the RNA polymerase and make it insensitive to regulation. Aslo, you can mutate RelA and make it inactive. And there are several other possible compensatory mutations... In order to notice these changes in the strain you made you really need to run a lot of tests, and the authors did not.

So here is another example how bacteria are cleverly trying to fool systems biology approach (another example is here).

Update: the SpoT knock-out strain used in the original paper indeed was iffy, it had compensatory mutations in RelA and an erratum was published, which I discuss here.

References:

Shachrai I, Zaslaver A, Alon U, & Dekel E (2010). Cost of unneeded proteins in E. coli is reduced after several generations in exponential growth. Molecular cell, 38 (5), 758-67 PMID: 20434381

Potrykus K, & Cashel M (2008). (p)ppGpp: still magical? Annual review of microbiology, 62, 35-51 PMID: 18454629

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

Plata G, Gottesman ME, & Vitkup D (2010). The rate of the molecular clock and the cost of gratuitous protein synthesis. Genome biology, 11 (9) PMID: 20920270

Mendeley group on stringent response

GFP-based quantification of proteins in the cell - don't try that at home!

This is something that come out of our tracking experiments in J. Elf lab. I guess we will never publish any quantitive account of it, so it guess it constitutes perfect blogging material.

GFP and its derivatives are widely used to label and track proteins in live cells. This way people follow localization of the protein of interest and accessing changes its concentration.

And here is where the fun begins. GFP has a fluorofor that is bringt only in a) oxidized b) deprotonized state. Therefore redox potential and pH of the cellular environment have profound effect on the brightness of GFP. Cultural media per se can change GFP behavior dramatically, and this is something outside of the cell...

Another problem with GFP is that it can radically affect stability of the fusion protein. We are not even talking about function here, we are talking about the number of molecules per se.

While struggling with our Dendra2 GFP variant and RelA_Dendra2 fusion in vivo we observed all these problems constantly. Unhappy cells were dark, with GFP in the dark state (pH? redox? you simply don't know!) and the numbers of proteins you detect were invariably much lower than what is estimated for the wt, un-tagged RelA.

But all this still does not stop brave souls from using GFP for single molecule quantification of bacterial proteins en masse. Whole library of YFP (yellow fluorescent protein) fusions was created and their copy number as well as diffusion characteristics were analyzed, and then systems biology happened to the dataset. No one knows what exactly does it all mean and how the numbers of detected YFP-fused proteins relate to numbers of the wt proteins, but this is details.

References:

Bogdanov AM, Bogdanova EA, Chudakov DM, Gorodnicheva TV, Lukyanov S, & Lukyanov KA (2009). Cell culture medium affects GFP photostability: a solution. Nature methods, 6 (12), 859-60 PMID: 19935837

Taniguchi Y, Choi PJ, Li GW, Chen H, Babu M, Hearn J, Emili A, & Xie XS (2010). Quantifying E. coli proteome and transcriptome with single-molecule sensitivity in single cells. Science (New York, N.Y.), 329 (5991), 533-8 PMID: 20671182

Brian P. English, Arash Sanamrad, Stoyan Tankov, Vasili Hauryliuk, & Johan Elf (2010). Tracking of individual freely diffusing fluorescent protein molecules in
the bacterial cytoplasm arXiv q-bio arXiv: 1003.2110v1