It's been a little while since I updated everyone on my long-running series of experiments, but I do have some news. At last report, I'd set up a large crucial experiment, and actually seemed to get it to work in the notorious Vial Thirty-Three. Of course, when you get something remarkable to happen in the lab, the first thing to do is see if you can make it happen again, and that's where the trouble started. The first repeat was rather nasty, and the second one was no better. I was baffled, since the first run had looked so promising.
My colleague Joanne, who was analyzing these samples for me, was puzzled, too, but she at least knew something to do about it. There's a huge benefit to working with people who know what they're doing. She took that third run and ran the samples for me again, this time in a much longer gradient on the LC/MS. (For the non-chemists out there, this means that the purification part of the method was extended, spreading out the various components of the mixture more). The control vials looked just like they had in the first run - not much, which is what controls are supposed to look like. The experimental vials had looked the same way, though, but with this new run my data appeared as if the results had come out from behind a cloud. Suddenly it looked like the first run again!
They'd been in there all along, as it turns out, and a cloud of ion suppression is what they were hidden by. This is a real problem with mass spec methods using mixtures of proteins (and the stuff that keeps them happy). There are a lot of reasons for this, only some of which are well understood, but having your analytes disappear and reappear unpredictably on you is apparently a widely shared experience.
I tried to see if there was some single component in my brew that I could leave out and thus fix my problem, but I should have saved my effort. That rarely seems to be successful - the real solution, as would have been clear to a real chemical biologist, is to run things the way you have to, and then clean up your samples before they go into the machine. The best way to do that is probably solid-phase extraction (SPE), which entails loaded your mixture onto some sort of powdered polymeric stuff which binds the analyte you care about. That lets you wash all the gunk out of the system, and then you use a different wash to elute the good stuff.
Here's an older review that illustrates the principle. These days, there are dozens of competing SPE technologies from all the major lab vendors. I evaluated a set of the more popular ones by setting up a row of dummy experiments - all my proteinaceous stuff, spiked with a constant amount of my desired product. All of them improved things, but one in particular (the Waters Oasis MCX, for those curious) seemed to do the best job, although I'm sure that there are others that would work as well. The method I worked out for it was the most complicated of all the ones I tried, but it's probably washing out the most sludge, too, because I'm getting ten- to twenty-fold more signal than I did before.
So, late last week I set up my first "Vial Thirty-Three" experiment again and worked it up with the SPE. It reproduced perfectly, to my great happiness, which takes me right back to the edge of things. Before I left the lab on Monday, I set up another run, this time with six different control and experimental arms, in duplicate, the most comprehensive look at this effect I've ever taken. I'm working it up today. Results in a couple of days, most likely. I'll keep everyone informed.