Directed evolution can be done faster and more effectively

Artificial (or directed) evolution experiments were for a long time to me some kind of easy to understand concept. It's clear indeed, you take a gene (resembles a genotype), mutate it or part of it in preferentially random way, express it (to obtain a phenotype) and choose ones that match you criteria. This process you can do again and again until you exhaust your passion.

In a huge contrast to the theoretical concept, things become much more hazy once you start designing directed evolution setup (I have never done it myself). Here might come problems: size of the library (the bigger the better, if you can experimentally afford it), how much time you can spend on doing this exercise (otherwise, how long each evolutionary round will take) and you also be considering possible biases that comes from the experimental system you use. For instance, in vitro systems that normally compartmentalize a single gene and its protein product  are not self-sustaining (see my another post how researchers are trying to make this system sustainable). Whereas using in vivo evolution methods you'll be applying a selective pressure on a whole organism rather than only on you gene of interest so you might end up with more sophisticated pattern of unintentional selective pressures. The latter for instance affects the most advanced method 'phage-assisted continuous evolution' (PACE, to read about in Nature journal or on HHMI). It is great in terms of its ease. It relies on a phage M13 propagation (this makes very short turnover time) and its all happen in one constantly replenished medium (so it continuous!) Also, due to the fast phage proliferation it will let you pick even the most rare but very efficient mutants of you gene.

Here are the E.coli cell infected with M13 phage. It lacks one important protein (pIII) who's gene is on the extra plasmid (red circle). You transform this plasmid along with a plasmid bearing gene of interest (green circle) that will drive pIII expression and will result in maturation of infection-competent phage.

Again, this system is limited since the obtaining of you most precious active mutant is dependent on the phage replication who's biochemistry is totally irrelevant to you.

So, it took only two years to design another directed evolution system that almost fully lacks this drawback, though it has some others (will tell later about this). It's been published in recent issue of Nature biotechnology journal. So, instead of using phage as a 'machine' to amplify the protein mutants group of Professor Andrew Ellington (here is the link to his lab) simply coupled production of your gene to the synthesis of Taq DNA polymerase that we normally use in PCR (they call it 'compartmentalized partnered replication' CPR, don't get confused). The design looks very similar to the PACE:

The only difference is that your drive the expression of the enzyme only (not a phage). Next, you insert you cell into vesicles along with primers and simply do a PCR. Thus, PCR will occur on those vesicles containing larger number of Taq polymerase molecules. Again, similarly to the PACE we dealing with amplification that lets you to pick up the most rare mutants. The problem here that it is discontinuous. Although, authors consider this beneficial as it ads another leverage to evolutionary process. Also, production of your library is still dependent on an organism that means you will not be able to produce toxic to E.coli proteins (this might be an issue only if production of Taq-pol requires heaps of you protein).  Thinking of both systems,  I might be not right,  but they do not have negative selection that may a be a problem as well. Anyway, the CPR system due it simplicity might be used  for adjusting not only a single gene, but rather a whole circuit. A big step ahead for Synthetic Biology!

Darwinian evolution in a primitive cell

Hi all. I feel totally ashamed because of my poor blog performance. I simply could not find a single hour to write a short review of papers that I have recently read or skimmed through.

This time I would like to discuss a recent paper of Professor Tetsuya Yomo from Osaka University. In this paper authors have established a platform for artificial evolution of systems containing of multiple components. As an object for selection they have chosen RNA directed RNA polymerase of Qbeta phage (Qbeta polymerase). This protein is known as the most productive RNA-polymerase and according to previous reports can produce up to 10¹⁰ copies of a single RNA template in 10 min. However, there are a major problems with this enzyme: (1) it's incredibly specific and can amplify only phage RNAs; and (2) it is often contaminated with phage RNA-derived small parasitic RNAs that effectively outcompete replication of genomic RNA. Thus whoever can 'teach' this protein to replicate RNA molecules of our interest that also will be resistible to the presence of parasitic RNAs can build (to say the least) very efficient way to amplify RNA molecules. 

What's more important you can design an artificial cell that will be contained of many components who's RNA templates will be replicated by Qbeta polymerase. Further application will be really dependent on you imagination. Firstly, you can explore the very basic principles of evolution of complex per-biological systems. Thinking of more practical and immediate use: one can design a self-sustained cell-free translation system, that will only be consisted of proteins and RNAs that you need and properties you dare to want. 

Alternatively, even without approaching such a tantalizing goal, you can perform an artificial evolution experiments of the systems of proteins. This was not possible so far and the protein engineering exercises have been done only for single relatively 'simple' protein species.

So what group of Professor Tetsuya Yomo have done? They simply overcome the problem of presence of parasitic RNAs by compartmentalizing the replication reaction in lipid vesicles. Such that non-functional RNAs that were prevailing in the replication mix were to a major extent eliminated during evolution iterations. Also, in order to be able to keep the mutant RNAs and Qbeta replicase isolated from the others mutants they fused the RNA replication vesicles with vesicles that contained E.coli translation extract (so called PURE system).

This compartmentalization approach reminded me a story of one of the prisoner's dilemma variants that was explained in 'Supercooperators' book, when 'defectors' that use resources of others without giving back normally outcompete 'cooperators' in a homogeneous environment but as soon as group of interacting 'units' become isolated - cooperators thrive. This, for instance, was a good way to show the importance of a group selection in evolution.

Although the authors have not reached the ultimate goal to have omni-reactive Qbeta polymerase (in fact they have not even mentioned this in the paper) they were able to get mutant polymerase that was efficiently replicating its RNA long template (mutant again) even in the presence of parasitic RNAs. That means, that authors, are well on the way to design a replicase that will be able amplify RNAs containing genes of our, or in fact, their interest.

Interestingly, that evolution of such primitive cell (as they call it) which is not alive demonstrated traits of Darwinian (i.e. biological) evolution. These are 'diminishing returns' (where the smaller benefits can be gained per mutation upon reaching the optimum) and more importantly 'rate of mutations' happened to be constant throughout all evolutionary experiment. The same have recently been shown for E.coli cells (living matter!). I wonder if this rate is close to the optimum of viruses or relatively simple bacterias.