Here's the gist of Derek's summary. We often talk about essential genes in an organism -- genes whose loss of function is incompatible with life under some defined circumstance. Genes like that are of great interest from multiple angles, especially if you are paid to look for drugs to kill disease-causing organisms, since interfering with an essential function should be a good way to do that.
One of the key molecules of life is pantothenate -- which we call vitamin B5 since humans can't make it. Indeed, all of us metazoans rely on other species to make this key molecule. Pantothenate is critical to fatty acid synthase (which we have) as well as things we don't have but take great advantage of such as polyketide synthases and non-ribosomal peptide synthases -- both important sources of valuable drugs. These systems are gigantic multi-domain proteins with multiple enzymatic active sites. Pantothenate plays the role of a flexible robot arm, swinging the substrate from one active site to the next. It's also a key component in making coenzyme A, which is critical for energy metabolism.
A key bit of the pathway is PanD (aspartate 1-carboxylase), which is required for making beta alanine which is a key upstream intermediate on the way to pantothenate. PanD is an essential gene in E.coli -- unless you feed the bugs beta alanine. Beta alanine isn't typically found in the environment, hence the need to synthesize.
So what the experiment did is first knock out the panD gene in a mutator strain and then ratchet down the ration of beta alanine in the medium, eventually eliminating it to select for any mutants that could still grow. And they found mutants that upregulated uracil production and there's a side-pathway that can lead to beta alanine via a toxic intermediate. Interestingly, the mutations were in repressors not structural genes for the pathway.
So then they deleted the key operon activated by the repressor mutations (rutABC) in the panD null strain -- and repeated the selection scheme. And found a mutation in yet another gene, speC, which reduces the specificity of the enzyme produced so that it can generate beta alanine.
Repeat with the triple mutant? Again E.coli found a route out -- though this one is apparently still being characterized. And we are left wondering just how much deeper does this go? If one makes the quadruple mutant, is there yet another pathway that pitches in? And another?
So for this one key metabolite there is apparently a lot of latent capability in E.coli to make it -- nature settled on one route but there were many other options. It may be that the known route was found first or could be that the tradeoffs for the other routes are much greater.
Obviously for antibiotic development this sort of knowledge is valuable -- if you are targeting key pathway X and it turns out there's a whole bullpen of alternatives, that is a bit of a concern. It might not be a show stopper -- the liabilities of the other pathways may still leave the pathogen sufficiently weakened -- but it might well be a problem.
There's a number of interesting additional directions one could take this, given sufficient resources -- which, sadly, I don't have, let alone the excessive resources required for what I'll outline below.
One would be to ask how this would play out in another model system. Suppose the experiment is run in B.subtilis? Would the same set of backups and alternatives emerge, or different ones? Or S.cerevisiae? Are there other pathways in other organisms which are primed to take on this role?
Which leads to another whole direction -- how many more essential genes could be played with in this manner? The key bit is one must be able to supply the key metabolite externally. That expands the playing field to a very large number of genes -- though not every gene.
But perhaps every gene could be in play -- with appropriate genetic trickery. If one has a very good controllable expression system, then the dosage of the essential gene could be controlled rather than the dose of a metabolite. However, in such a design the easiest way for life to evade one's designs would be to break the expression control system and release the essential gene from control. So that would need to be designed very carefully and probably with multiple controls. For example, two independently regulatable control elements might be placed on the gene (say a Tet repressor and a riboswitch) and the essential gene also in tandem with (or fused to) a visible reporter -- so that one can quickly discriminate against escapees.
The most interesting system to re-run this sort of in would be JCVI's minimal genome -- which lacks panD (I looked!). Craig Venter and colleagues claimed they had whittled down the already minimal Mycoplasma genetic complement to about 450 genes and could go no further. But that just asked what could or could not be removed without given the bug any time to adapt to that deficiency. Could a panD sort of experiment shave a few more genes out?
One other thought: has nature ever resorted to any of these alternative pathways in place of PanD? If one searches the myriad of fully sequenced bacterial genomes, which are missing the canonical pathway? And if so, are they all actually auxotrophs or can we find examples that have gone to one of the backups? That would require some guessing about the environment -- after all, Mycoplasmas apparently swipe panthothenate from their hosts (the SP4 medium used to grow them is insanely rich, with yeast extract and beef heart infusion). Just running the search could be a very interesting advanced undergraduate or rotation student project.
Alas, I don't have regular inflows of impressionable students to direct into exploring these. If anyone who does have such resources ever explores this, I'd love to hear about it.