Chasing Down the Cause of Random Experimental Results

I'm pretty sure that every wet lab biologist has a story or two or many about experimental controls behaving weirdly or in seemingly unexplainable ways. A very small percentage of the time, chasing down the underlying cause can actually lead to a Nobel prize. Much more often the underlying cause is unremarkable ("Oh, that was a ug instead of mg?"). We had one of these results in the lab last week and, while interesting and actually a real phenomenon rather than facepalmish error, it's still fairly unremarkable so I figure I'd share it here.

We are currently trying to measure mutation rates to rifampicin resistance in Pseudomonas stutzeri. We already have encouraging results showing how a particular genotypic change increases mutation rates to streptomycin resistance and to goal is to calculate mutation rates an additional phenotype so we could begin to think that the genotypic change generally increases mutation rates. For other reasons, we wanted to measure the mutation rates to rifampicin resistance in a streptomycin resistant background strain.

Since P. stutzeri is competent for natural transformation, we transformed a rif and strep sensitive strain (rifS and strepS hereafter) using genomic DNA from a rif and strep resistant strain (rifR and strepR). No problem yet as we got plenty of strepR colonies back.

The next step is to use a fluctuation test to calculate mutation rates to rifR using cultures started from a single strepR colony. The basic idea of a fluctuation test is to grow many independent cultures starting from very low cell densities (typically ~1000 cells, in order to insure that there are no rifR colonies at the start), and then plate the entirety of these cultures under selective conditions once the cultures have grown to appreciable densities. If there are no rifR cells at the beginning of growth , you can use the distribution of rifR colonies that appear across the independent cultures in order to calculate mutation rates.

The picture below describes what you might expect from a typical fluctuation test.

When you plate out samples at time 0, there should be no rifR colonies (which I'm showing as blue circles if present). After growth, in this case 24 hours, rifR cells will have arisen by spontaneous mutation INDEPENDENTLY in each culture. As you can see from the picture, the expectation is that some cultures will not contain any rifR cells, some will have a few, and some will contain many rifR colonies. Those with many colonies are called "jackpots"and may appear as confluent lawns, hence the totality of blue in the picture. There is a distribution of rifR cells across independent cultures because mutations that lead to rifampicin resistance will occur at different points of the growth curve in each culture. A single mutation that arises during the first cell division after starting the experiment will lead to jackpots because these rifR cells have the chance to divide many times before plating. A rifR cell that arises during the last division before plating will only be represented by one colony because it hasn't had the time or resources to divide and proliferate. You can use this distribution to actually figure out mutation rates. Here is Stan Maloy's explanation of the fluctuation test.

Back to our P. stutzeri experiments...when we plated out the fluctuation test for our (what we thought) strepR rifS isolate, we got this result back from the time zero plate.

There was a lawn of rifR bacteria, which is statistically higher than zero (not really, but go with me on this). Even though the mutation rate to rifR is relatively high, something was fishy here because there should have been no colonies. We went back to the original plate of strepR transformants, and streaked additional isolates to both strep and rif plates. Even though the strain which we had originally transformed was rifS, I was surprised when roughly half of the strepR transformants were also rifR.

What could explain this result? The mutations that give rise to the strepR phenotype usually occur within a gene called rpsL. This gene codes for one of the many proteins that make up bacterial ribosomes, and indeed, the mechanism of action for streptomycin is inhibition of translation. The first thing I did was investigate the genomic context, what genes surround rpsL, within the one previously sequenced P. stutzeri genome. Here's what I found at the Pseudomonas genome database (which is extremely handy BTW).

All annotated genes in this section of the P. stutzeri genome are represented as red boxes. rpsL is the red box surrounded by a very dark line. The numbers on the bottom of the picture display the position of each gene in the P. stutzeri genome (rpsL is roughly at position 888,000 out of ~4,000,000). Here's where it gets interesting. Mutations that lead to rifampicin resistance typically occur in a gene called rpoB. This codes for a subunit of RNA polymerase, which makes sense (again) because the mechanism of action for rifampicin is to inhibit transcription. As I described above, we originally made the strepR strain by transforming a rifS strepS strain with genomic DNA from a rifR strepR strain. rpoB and rpsL are only about 5000 bp apart and I originally had no clue that rpoB and rpsL were genomic neighbors. Since, during natural transformation, genomic fragments sized 5Kb and above can be recombined into the recipient genome I'm guessing that a substantial fraction of the cells recombined both strepR and rifR mutations during this step from the donor genomic DNA. Since the exact positions of recombination are essentially random, cells that remained rifS after transformation must have recombined smaller fragments that only contained the strepR mutation. Here's a quick picture of what I think happened.

Rifampicin resistance is colored red, streptomycin resistance is colored blue. Cells that are rifR strepR are colored purple. While there were likely rifR strepS cells generated after transformation, these won't grow on the strepR plate so that's why there aren't red colonies. Mystery solved...and this maybe something I can utilize in future experiments although as of right now I have no idea how. If we would have originally picked a rifS strepR colony for the fluctuation test, I wouldn't have realized any of this. Sometimes science is random.

How I came to work with plant pathogens

It's getting to be grant season for me again, and one of the most important parts about writing grants is learning how to sell and justify your research story to a larger (somewhat less specialized) audience. I've also seen some stirrings on blogs and twitter about how to choose postdoc labs and research programs in order to set up an academic career. Although I think that the best answers to the postdoc question are much more dependent on both the individual and the lab, both of these topics inspired me to jot down some ramblings about how my own research program developed.

I've always been interested in understanding how microbial populations adapt to new environments and my graduate school career was spent studying evolutionary dynamics of the human pathogen Helicobacter pylori. Dr. Karen Guillemin had just started her lab at the University of Oregon and I'm most thankful that she was willing to take a chance on having an evolutionary biologist play around in her lab, which at the time was much more focused on understanding the molecular biology of bacterial pathogenesis. I was lucky to be co-advised by Dr. Patrick Phillips (a nematode guy), to fill in the blanks when it came to evolutionary biology, population genetics, and random Star Wars references. My grad school experiments centered around setting up a laboratory evolution system, modeled around the wonderful work of Rich Lenski, where I could test for the effects of genetic exchange on rates of adaptation. I'll talk about the ins and outs of those experiments in a future post, but at the end of graduate school I was left with a huge choice as to where to do my postdoc.At this time experimental evolution was exploding as a research field, and I wanted to continue studying bacterial adaptation using experimental evolution, but I also wanted to begin to study populations within hosts. I actually had a great interview with Rich Lenski in February 2006 in East Lansing where we talked about teaming up with Jeff Gordon at WashU to look at the effects of Rich's E. coli laboratory mutations during mouse infections (among other things including college basketball). After that interview, I was almost convinced that I was going to become a Michigan state Spartan.

There was a counter-voice in the back of my head, however, that maybe I should jump a little bit more out of my experimental comfort zone. I was feeling leery of working with mammalian pathogens for a variety of reasons: 1) I'm a fan of cute fuzzy animals like ferrets, and it would be very difficult for me to actually do research within hosts 2) Coming out of the Phillips lab I was well schooled in the importance of sample size for statistics. It's very expensive to perform large numbers of infections in anything with 4 legs and hair, and I was worried about getting high enough numbers of replicates to actually make sense of evolutionary trends. I knew I wanted to continue on in academia, and I had a serious internal discussion about how easy it would be to fund a lab that performed the experiments with mice that I was imagining. 3) I had a sneaky suspicion that genomics was about to explode (indeed, it already was in 2006), and I wanted to jump on the train. I certainly could have done this in Rich's lab but even at that point the field of the genomics of mammalian pathogens was getting crowded.

So what was my other option? I miraculously came across a random paper out of Jeff Dangl's lab, and had never heard of Dangl before this point (looking back...the only name I recognized on the paper was Dave Guttman from his work on recombination). I shot a quick email off to JD laying out my interests and and was pleasantly surprised that he was willing to fly me out for an interview in Chapel Hill. This interview was scheduled to be a couple of weeks after my interview in Michigan (in February). While can't say that the contrast in weather consciously affected my decision, it was 20 degrees and snowy in Michigan and 70 degrees and sunny in North Carolina during my visits. Dangl was well known in plant biology circles at that time for helping to work out the genetics of plant immune responses to pathogens, indeed, he was elected to the National Academy a couple of years later. Jeff was embarking on a pretty ambitious (as with seemingly all Dangl lab research) project to use "next-generation" sequencing technologies to illuminate genomic diversity in phylogenetically divergent strains of the plant pathogen Pseudomonas syringae. I knew that P. syringae was related to Pseudomonas fluorescens, which is a popular system for experimental evolution studies thanks to Paul Rainey and colleagues, and I figured that I might be able to piggy-back off of that system to set up my own P. syringae evolution experiments one day. This has actually proven to be spot on:) A postdoc at UNC would also throw me into the fire of bacterial genomics and force me to learn how to sequence and assemble strains, perform genomic level experiments, and PROGRAM! Perhaps most importantly, plants grow in dirt (which is dirt cheap) and no one really cares how or how many plants you euthanize during experiments so long as any transgenes stay contained. Working with plant pathogens would enable me to carry out an appropriate number of replicates to try and silence the subtle voice of Patrick Phillips that remains in my head today when I think about statistics and experimental design.

One other helpful piece of evidence that sealed the deal for me to join the Dangl lab was learning that mammalian and plant pathogens pretty much deploy the same sets of tools during pathogenesis. Evolutionary patterns from one system are very similar to the other and research findings are relevant across both, I wasn't really going to miss much studying plant pathogens instead of Shigella. For instance,  the importance of type III secretion systems during infection was actually recognized at about the same time for both Yersinia pestis and P. syringae! The similarities have been laid out clearly in paper form a couple of times (here, here, here, here, here, among others) but the main differences between systems arise when immune responses in the hosts are compared. Even then, innate immunity is still a major shared component that contributes significantly to defense responses. As such I'm able to apply for grants across the main funding agencies (NSF, USDA, NIH), which theoretically makes make my life as a PI a little bit easier, although it hasn't yet:)

To be completely honest with you, I was torn after visiting both labs and took a couple of weeks to make my decision. What finally did it for me was flipping a coin, not once or twice, but until there was a run of the same side. UNC came up as the answer about 5 times in a row, and I found myself completely OK with that decision. Sometimes you just have to jump in and not look back.