Tuesday, February 21, 2006

The realities of lab work

As I start to do more labwork, I'm gaining an appreciation for two things:

- The often-peddled view of "breakthrough" experiments obscures the fact that the majority of time needed to do experimental work is spent just making all the stuff you need for the experiment you really want to do. Doing the "final"/"breakthrough" experiment itself is often one of the quickest bits.
- A lot of experimental molecular biology is based on exceedingly unlikely events and making it up on volume.

Let's take the second bit first. An analogy, inspired somewhat by personal experience: suppose you have your windows open on a warm summer night and a bat flies in and starts flapping around in circles in your living room. You'd like to catch the bat and toss it back out on its ear, so to speak. One [admittedly batty, haha] scheme would be to reason as follows: a bat is sort of like a mouse with wings. Now, what catches mice ? Cats, that's what. Are there any cats on hand ? Yes, there are. So, how about tossing a cat into the air and hoping that it catches the bat ? [Assume, for pedagogical purposes, that the cat available to you is sufficiently composed to recognize the most important aspect of the situation, viz. that there is a mouse to be caught, and ignores such trivialities as the fact that it's being tossed into the air.]

Now, the chances of the cat catching the bat on any single toss are miniscule. However, if you toss the cat, say, a million times, the chances that the bat is caught start to rise into the realm of not-totally-improbable. And that's how a lot of experimental procedures in molecular biology work: you throw together a bunch of things that might snap together in the right way maybe once in a million times, but if you do this with bajillions of cells [and you can easily get a million cells in a milliliter of cell culture], the chances are good that you'll get a few instances where things snap together correctly.

Ok, so you get lucky occasionally, but given that lots of lab work consists of mixing colourless liquids together to make more colourless liquids and you can't really look inside cells to see what's going on at the molecular level [recent cool developments notwithstanding], how do you figure out when the desired event has taken place ? This is where the high art [and science, I suppose] of genetic selections comes in. A lot of genetic selections are predicated on arranging things such that if the right thing happened, the critter lives; if the wrong thing happened, the critter dies [or vice versa; both types of selections are used].

For example, a commonly used selection is antibiotic resistance: take cells that are normally vulnerable to an antibiotic like ampicillin, perform your experimental procedure on them and then looks for cells that have acquired resistance to ampicillin. If the right thing happened, a small fraction of the cells will have acquired a gene that makes them resistant to ampicillin and will grow on an ampicillin-containing medium whereas the rest of the cells that you used for your experiment will die. Genes like the ampicillin resistance gene are referred to as "markers", because they mark the cells that you want.

A concrete example of all this: one of the ways to insert foreign DNA into a yeast cell is to use a mechanism called homologous recombination. Pretend the yeast genome consists of actual text, like:

"Tonight I ask you to pass legislation to prohibit creating human-animal hybrids because I think they're the work of the devil"

If you now introduce a piece of DNA whose beginning and end are the same as some stretch of genomic DNA into the yeast cell, like:

"creating human-animal hybrids like me and other members of my administration because"

then, a few times in a million, there's a bit of "cut and paste" so the yeast genome becomes:

"Tonight I ask you to pass legislation to prohibit creating human-animal hybrids like me and other members of my administration because I think they're the work of the devil"


This sort of swapping is called homologous recombination. It's a rare event -- the new piece of DNA needs to "find" a stretch of genomic DNA that matches at the right spots and, even for something as simple as yeast, the genome is over 13 million letters long so there's lots of ground to cover; the cell is a pretty crowded place; the right protein machinery needs to be around to make the swap happen etc. And, of course, you have to get the details of the experimental procedure [mostly] right.

Even though it happens so rarely, homologous recombination is one of the main ways used to modify the yeast genome. The selection that's often used to find the cells in which it happened is to use cells that can't make an amino acid that they need [and so they won't grow on medium without that amino acid] and, together with the new DNA, introduce a gene that allows them to make that amino acid. Cells that have successfully recombined the new DNA into their genome then acquire the ability to grow on medium lacking that essential amino acid, which ends up looking like this -- the white spots are colonies of yeast cells that are able to grow successfully.

Of course, there's now another complication: the number of easily usable marker genes is fairly small, certainly less than 10 for yeast. If you use up one marker gene each time you modify a cell, this means that you're limited to less than 10 successive modifications. So if you want to make lots of changes, you have to figure out a way to recycle your marker genes. Doing so often requires that you have what's called a "counterselectable" marker -- if you grow cells in one way, only cells that contain the marker survive; if you grow them in a different way, only cells that don't have that marker survive. So the entire procedure requires that you first do your experiment such that only cells in which the right thing happens acquire the marker and survive and you then take the survivors from the first step and grow them under different conditions so that only the ones that have now lost the marker stay alive.

Each such iteration takes, say, 1-2 weeks if you're good at it, have all the necessary raw materials [like the pieces of new DNA] readily available, and don't run into any trouble [which is, unfortunately, not a sufficiently rare event ...]. So, to make, say 10 changes in the genomic DNA of a yeast cell, you're talking 3-6 months until you have the yeast cells that you want to actually experiment on.
If, along the way, you've had to construct the new pieces of DNA yourself, a process that's pretty much the equivalent of self-flagellation, it'll probably take twice as long. Throw in a fudge factor of "I've never done this before, so I'll make lots of mistakes along the way" and you're up to about a year.

Now you can finally start doing things like checking how these cells respond to being poked with the molecular equivalent of a sharp stick. And, of course, if you're unlucky, they won't do anything interesting at all, so you've just spent a year of your life on mind-numbing work and constructed something that's about as exciting to watch as this. Congratulations !

All this is just a long-winded way of saying: this whole lab work thing is definitely 99% perspiration, 1% inspiration.

2 Comments:

Blogger Son2 said...

Wow.

11:20 AM  
Blogger Corey said...

Well written. I like reading what you do in lay terms that my wee brain can understand!

5:46 PM  

Post a Comment

<< Home