Selection-driven Gene Loss in Bacteria

Bacterial genomes differ dramatically in size: from 140Kb to 13Mb (those numbers might be off now...please let me know if something has broken the record. Yes, I know the lower estimate can change based on semantics, but there are a bunch in that range). Although we have some clues as to how selection acts (or fails to act) on genome size, outside of intracellular parasites it's a bit of a mystery how selection shapes total genomic content. Perhaps the most interesting case out there involves genome streamlining in marine bacteria, which has been attributed to selection but which remains a just so story to this point.

One easy explanation is that "extra" DNA is costly in and of itself because it takes things like carbon, phosphorous, and nitrogen to physically make DNA. While probably true in the strictest sense, as far as I know there has not been a very clear test of the actual selective forces that act at this level. I would guess, especially given what's come out of the arsenic life debacle, that bacterial cells can survive just fine the way they are with low levels of phosphorous et al. I'm not sure how many environments are limiting enough for these elements to have direct selective effects on genome size (although see situations like this, this, and this). Other recent research points to the "cost" of extra DNA residing in the production of RNA and proteins. In this experiment, proteins and ribonucleotides are not inherently costly (*under the environments tested), but production of unnecessary proteins likely takes away cellular machinery that could otherwise be put to better use. There are only so many ribosomes in a cell to carry out translation. If these are occupied by unnecessary transcripts, they can't be used to produce more essential proteins. Interestingly, these costs may change based on previous environment. There are also some additional other hypotheses for genome size evolution that I may touch on in the future, but for now I would like to give a brief overview and thoughts about a paper relevant to this question that came out last week in PLoS Genetics from Dan Andersson's group.   

Schematic of how they isolated deletion mutants

This manuscript is basically laid out in three related, but independent parts. The first consists of measuring the rate of deletions throughout the Salmonella enterica var typhimurium LT2 chromosome (see figure above). They first hopped a transposon containing three phenotypic markers into random areas of the chromosome. The markers contribute 1) resistance to chloramphenicol 2) cleavage of B-galactosidase leading to blue colonies during growth on X-gal 3) sensitivity to chlorate. They can measure rates of deletion using this transposon because of a cool genetic trick: the moaA marker renders the cells sensitive to chlorate, so they can select on resistance to chlorate in order to identify when the transposon might have been deleted from the chromosome. Once they get these chlorate resistance mutants, they look for white colonies and those that are sensitive to chloramphenicol in order to eliminate cells that have only deleted small portions of the transposon or have inactivated moaA through mutation. 

The "deletometer" transposon

They use this "deletometer" to measure rates of mutation in 11 chromosomal regions and find that this rate varies by 2 orders of magnitude. These deletions range up to 10's of thousands of bp in size. Kind of a sidenotes to the total story here, but they do find evidence for the existence of a RecA independent deletion mechanism in S. enterica by studying genomic context of the deletions (RecA needs 25bp of sequence similarity to recombine pieces of DNA, but they find evidence that there is much less similarity bordering many of their sampled deletions, and in some cases none). They also find that, as shown for other bacteria like Bordetalla, that the replication terminus seems to be a hotspot for deletions (higher rates at terminus).  The one question in my mind that remains from this portion of the paper is how they control for genomic context. It seems like regions of the chromosome that have more redundant sequence should have higher deletion rates, but maybe I haven't thought through this enough. 

 Fitness effects of deletions across enviroments and assays 

Next they use a subset isolates to test for fitness effects of the deletions (and therefore address the question of cost of extra DNA). By measuring growth of strains in two environments (rich and minimal media) they show that some of the deletions actually increase growth rate in 15 of 55 cases. They further reinforce that deletions can be beneficial using growth assays where two strains are directly competed against one another. Importantly, they find no relationship between the size of deletion and the fitness effects (strike for the DNA is costly in and of itself camp).

The third part of the manuscript consists of a 1000 generation passage experiment in rich media. Over the course of laboratory passage and adaptation, there are a suite of deletions that reach high frequencies and are therefore likely adaptive during lab passage and in the right genomic context. They recreate these deletions in an unadapted chromosome and are able to show that 2 out of 6 do increase bacterial fitness in the lab (the absence of effects for the remaining four they chalk up to epistasis...they are beneficial only in the presence of other mutations that are not present in the unadapted ancestral strain). 

The overall, memorable, take home message from this paper is that random deletions can be beneficial under some circumstances. Although this has been previously seen, this paper extends the result. They don't test the mechanistic underpinnings of these selective effects, which is what I hoped might be in the paper given the title. Although the authors don't talk about this too much, there a lot of deletions that are detrimental under each condition. The data seems to indicate that specific regions of the chromosome are more costly than others (beneficial fitness effects are only found in a subset of chromosomal positions), and I'm curious whether there is some unifying theme to these regions. Maybe they contain highly expressed but unnecessary, and therefore wasteful, genes. I'm curious why there are differential fitness effects for mutations that affect the same region...seems like it would be straightforward to figure out what differs amongst these different deletions within the same region as a way to get at the cost of DNA question. There is some additional novelty in showing that deletion rate varies over the chromosome, that there seems to be a RecA independent deletion mechanism in S. enterica, and that there is a deletion hotspot in the terminus. 


Here's the citation:


Koskiniemi S, Sun S, Berg OS, Andersson DI. 2012 "Selection-Driven Gene Loss in Bacteria". PLoS Genetics.