Last week, I introduced readers to a cell wall project. Recapping briefly, Joseph Hill (at Penn State) and I are asking whether a group of mutants have difficulty in organizing the secondary cell wall. I can look at cell wall organization with scanning electron microscopy (check out Fig. 1 from the Jan 3rd post). Joe grew the plants at Penn state and then cut sections of their inflorescence stem and sent them to me. Joe cuts sections on a cryostat; that is to say, a stem is sectioned while frozen. The frozen state acts a bit like embedment to make the stem nearly solid and thus section cleanly. The samples arrived early in the week and I prepped them up for microscopy.
Today, as I walked into the lab to finish the preparation, I thought about the many times over the years that I have headed labwards for my samples, pushing them through some protocol that extended into the weekend or night. At a time like this, it is easy to think of ‘my samples’ as ‘my children’, while I fuss over them, protecting them from contamination or from getting too hot or too cold.
Although an easy simile, a misleading one because ‘my children’ is a finite set whereas ‘my samples’ (usually) represents an infinite set. In the case at hand, the mutant genotype is known exactly (in principle, ignoring the possibility of additional, random, mutations), but we seek to know the consequences of the mutation for the plant. This is a phenotype, and phenotypes not only reflect the ‘instructions’ in the genes but they also integrate every random molecular collision, every cosmic ray absorption, every minute variation in soil, light, or wind. That is why two plants with the same genotype grown side-by-side always look different from each other. Which plant captures the effect of the mutation? Well both! How about a 100 plants grown side by side? All 100 of them.
One might extrapolate, then, that assessing the consequence of a mutation truly requires assaying an infinity of plants. There is a kind of Zeno’s paradox here because as we look at more and more individuals we intuitively think that the amount of un-sampled variation would go down (if 6 peas in a pod look alike, how about 6 million?). But in an actual infinity of peas, one would be a watermelon – every variation permitted by the laws of physics would occur. Infinity is big.
The resolution of this paradox is to ignore rare variations. If a “watermelon” pea occurs once in ten to the quintillion peas, we write it off as being of no interest. This allows me to get on with experiments but it is, or should be, unsettling. How rare is rare enough to ignore? One in ten to the quintillion is a ridiculous asymptote because the realistic limit is more like one in a hundred (experiments with 1000 individuals per treatment are rare indeed). And as I will describe below, for the experiments with Joseph, we are going to use a measly four individuals per treatment. We have little choice because the assay is expensive and difficult. Had we looked at plant number five we might have seen further variation. This is a caution that often goes unheeded as we think of our samples as our children, precious and complete.
Turning to the project at hand, given that each datum will arise from a tiny fraction of a cell’s surface area (the SEM images are about 2 µm x 2 µm), variation could be expected among regions of a cell, among cells, and among plants. I am willing to ignore the variation among cellular regions because anecdotally it seems smaller than among cells or plants, but is also surprisingly difficult to handle: opposite regions of a single cell are rarely both in focus.
One might imagine Joe sending me sections from say a dozen plants for each of the genotypes we want to check. Because there are five genotypes (the wild type and four mutant lines), that would be sixty tubes of sections (maybe six sections per tube). This would take me about a year to examine, and cost thousands of dollars of beam time on the microscope. We have to get by with fewer samples.
A practical limit, based on my time and cash flow, is four plants. That would add up to 20 tubes (four plants x five genotypes). To cut back on tubes in the mail, Joe made sample pools where he put a section from each of the four plants into one tube. So tube “A1” contains one section each from four plants of the A genotype; tube “B1” has one section each from four plants of the B genotype, and so on. He sent me three of these pool sets, so 15 tubes.
By the way, he really did label them “A”, “B”, etc., so I have no idea which genotype is which. Might help ease confirmation bias, though I have my suspicions about which is the wild type, because the “C” sections held together better than those of the other letters.
I processed pools 1 and 2 this week. By using small vials, I processed five at a time. Processing involves exchanging the ethanol for t-butanol, giving them fresh t-butanol three or four times separated by several hours, and then freeze drying the t-butanol to dry the sections). Today, I put sections on stubs, with 1 section of all five genotypes on each stub. That came out to eight stubs, with five sections each. By having each genotype on each stub, the coating process will be the same for all of them. Of course, I have to be careful that I can identify which section corresponds to which genotype. I cut off a corner of the mounting tape so it is asymmetrical and drew guides in my notebook.
My goal will be to image 25 cells per section and four sections per genotype, for 100 cells total, per genotype. This will sample the variation between cells well but the cells will come from only four different plants. In principle, I can do this with just the first four of the stubs but because some of the sections cracked up, I mounted more so that I can be sure to get 100 cells. In fact I will also process the third pool to hold in reserve. Over the coming weeks, I will image the material and then run some algorithms to quantify the organization of the cell wall. Stay tuned!