The Longest Quick Experiment of my Life
On the fourth page of this article is a sentence – “Hence, we replaced the HMR-E variant (AAACCCATAAC) with the genome-wide consensus sequence for the Rap1 protein (ACACCCATACATT).” This sentence is three years of my life.
The experiment should have taken two weeks. Okay, since I was a computational biologist who never held a pipette until coming to grad. school, I expected it to take a month or two. And my collaborator Bilge did the hard part of creating the construct; she actually did the tricky work of changing AAACCCATAAC to ACACCCATACATT. Now, all I needed to do was to pop that construct into the right place. Just a homologous recombination; in yeast, even a computational biologist should be able to do that!
Throughout the first year of my endless attempts, I just blamed myself. After all, when I came to the lab, I spent 30 minutes trying to turn on the Bunsen burner. I knew it worked as I had just seen the professor Jasper use it. I knew that if I asked anyone for help on this, I would be excommunicated from the lab and possibly Berkeley. I changed the tubing. I got another burner. Nothing. I was trying for so long, I was afraid of blowing up the lab. But I did not smell any gas, and that’s when dawned on me that I was switching handle between two “off” positions. Who knew that “on” is the middle setting? And I also used to call the 1.5ml Eppendorfs, “Beckendorf” tubes, in honor of the Berkeley professor Steve Beckendorf.
So it was entirely plausible that, as Russians say, my hands grow out of my arse, and I was simply not wired for experimental biology. Luckily, I had another project and could do some computational work to give me an illusion of progress, and this Rap1 endeavor was something I could take breaks from. By my third year of grad. school, I had succeeded in enough experiments and transformations to confidently say that something was definitely wrong at this locus. I had no problems doing transformations in other species, at other sites, with longer and shorter constructs – I just couldn’t get this particular sequence into this particular location.
Once or twice a year, I would complain about it again at a lab meeting, and kind labmates would give me 5-10 suggestions. Every month or two, I would try the experiment again, with variations based on the suggestions. Nothing helped. Finally, after three years, I got lucky. At the Toronto yeast conference, I had breakfast with Gilles Fischer, and after hearing my woes he mentioned that they had similar troubles and offered a solution. When I returned to Berkeley and tried his approach, the experiment finally did take the two weeks I thought it should. And that is the point when I was actually able to begin the experiments to answer the questions I had in the first place (four more years before the paper finally got published, by the way).
P.S. For the 4-5 other people in this world that may experience this same problem, here are the technical details. The problem was with the counter-transformation. Inserting the URA3 casette into HMR worked like a charm. Popping it out is what killed me. I simply never got colonies on the 5FOA plates. Though I did everything by the book and plated onto YPD plates after the transformation itself and replica-plated onto the FOA plates after a few days, nothing ever came up.
The suggestion from Gilles Fischer was to replica-plate from the YPD five or six times, again onto YPD plates, discard them, and then replica-plate onto the FOA plate. This sounds crazy at first, but I have a theory.
The URA3 casette was driven by a Rap1 promoter. HMR is near the telomere on chromosome III, and there is a ton of Rap1 in the vicinity. It may be that URA3 was over-expressed at this locus, and the cells in this strain had considerably more Ura3 than usual. Ura3, of course, is toxic on 5FOA plates, and it also has a lag. It takes a few generations to clear Ura3 from the cells, before one can see growth under selection. That’s why we first plate the cells on YPD after the transformation – to allow the dilution to happen. It may be that with extra Ura3, there was not enough dilution, and when replica-plated onto 5FOA, the residual Ura3 was toxic.
Why then did Gille’s suggestion help? Consider the colonies on the YPD plates, post-transformation. The colonies are three-dimensional and at the top are the youngest cells (with the least Ura3) and at the bottom are the oldest. When we replica-plate, we invert the order of young and old cells. On the top of the velvet are now the oldest cells, with the most Ura3. By transferring to the 5 extra YPD plates, we slice off sections from the 3D colony, discarding them, and getting to the lower layers with the youngest cells.
Of course, I never tested any of this. Was URA3 over-expressed at HMR? Would a different URA3-cassette, driven by a different promoter, work better? If correct, replica-plating from YPD onto 6 different 5FOA plates should show more and more colonies on each sequential 5FOA plate. I would love to see someone do these experiments!