Today, interestingly enough, we had a journal club meeting whose topic further supported the theory that random mutations can indeed result in extra functionality not present in the original strain.
This time, however, we need not even use "natural selection", but rely purely on several rounds of random mutagenesis combined with a selective assay.

The system in question is that of inteins, which are self-splicing proteins. Unlike RNA splicing, proteins that contain intein regions are able to modify themselves posttranslationally without the assistance of any external chaperones or factors.

All that is required is that the proper sequence be present, and that the N and C termini of the intein be close enough in proximity to catalyze it's excision. (See diagram below)


The current "problem" with inteins is that this process of self-excision is completely unregulated. As soon as the parent protein has folded, the process of intein excision is fairly straightforward. Apart from drastic pH or environmental cues, the excision process towards the final product is quite efficient.

To see if by some clever protein engineering a regulateable intein couldn't be manufactured, David Liu's group at Harvard thought up putting a small-molecule binding motif in between the inteins' catalytically active ends.
In a fairly groundbreaking PNAS [paper], Liu's lab was able to a technique called "directed evolution" to generate an exceptionally efficient system that used the small molecule 4-hydroxytamoxifen (referred to in the paper as "4-HT") to control intein splicing.

This was performed by taking the RecA intein, and replacing the homing endonuclease domain with a domain that tightly binds the 4-HT molecule, "ER LBD" (ER Ligand Binding Domain). When the ER LBD domain binds 4-HT, a structural change occurs that brings the N and C termini of the intein into close proximity, thus increasing the likelihood of its self excision. They then put this (N)-intein-ER LBD-intein-(C) construct into a gene that fluoresces in ultraviolet light, GFP.
The idea is that if the intein is present, it disrupts the GFP protein's structure, preventing it from fluorescing. If the intein excises itself, GFP's ability to fluoresce is restored. This allows for a quick and straightforward analysis of the inteins' excision state.

Once this (N)-GFPa-intein-ER LBD-intein-GFPb-(C) construct was prepared, it was submitted to a process called error-prone PCR which purposefully introduces mistakes into the sequence. Once this was performed, the sequence was transfected into cells in the presence of 4-HT. Only colonies that glowed green under UV light (i.e. cells that inteins which had successfully excised themselves) were chosen. (See diagram to the right)

This process was repeated until a mutated construct was found that:

A) Did not have considerable "bleed through" (i.e. didn't have inteins that excised themselves without the presence of 4-HT
B) Had efficient conversion of the full-length into the modified (shortened) version.

On this latter point was the project's only problem. Even at 1uM 4-HT, no more than roughly 50% of the full-length intein construct was converted into the spliced product.
Still, the creation of a completely new form of controllable postranslational modification is revolutionary.

In this example, "directed evolution" refers to the propagation of change in genetic material selected for consciously (usually by the researcher). This is opposed to natural selection, where changes in genetic material are selected for based upon how well that organism that sustained the changes compares in terms of viability towards its peers.
By correctly utilizing genetic modification techniques such as error-prone PCR, researchers are able to accelerate their research by doing several things:

First of all, they are able to quickly screen against mutations that are obviously detrimental (i.e. constructs that result in a toxic product). By choosing a reporter such as GFP, they are also able to quickly identify mutants that have the desired traits (whereas natural selection only selects for viability).
Furthermore, because error-prone PCR is completely random and non-specific, it can quickly cover many modifications to the protein that would otherwise be prohibitively difficult or time-consuming to perform deliberately. (It also results in non-intuitive solutions occasionally, but this is much less common).

So, the conclusion that one must come to is simple: Given a selecting force (whether it be the hand of the researcher, or cruel unyielding fate), random mutation can result in entire systems that display additional functionality beyond that seen in the simple sum of previously-existing parts.

For those wishing to read the entire article, Free full-text is available from PNAS [here].