Let’s say you’re a fisherman/woman trawling for tuna out on the azure-blue waters of the Pacific. Tuna’s your desired catch, but as you drag your net through the water you notice that all manner of aquatic life gets ensnared, to say nothing of styrofoam flotsam, plastic bottles, used automotive parts, and syringes. The FDA has guidelines about these sorts of things and the folks back at Trader Joe’s won’t tolerate even trace amounts of dolphin in their tuna. Bottom line is – you need your tuna to be pure. However, fishing individual tuna out of the sea one by one is extremely labor intensive, and though it may achieve high purity, you’ll be hard pressed to meet your production quotas.  The point of all this?

Scientists in the Nelson Lab at Brandeis fish for neurons. And not just any neurons, mind you, but very specific types. The end goal is to harvest their mRNA in order to “read out” their global gene expression using microarrays or sequencing based methods. They’re not alone in these pursuits; on the contrary, cell-type-specific gene expression profiling is a burgeoning field. However, like the analogy of fish in the sea, neurons exist in a complex and crowded environment, and isolating specific cell types requires some ingenuity. Different labs have used very different methods. In a recent study published in PLoS ONE, Okaty et al. compiled and re-analyzed all of the publicly available mouse brain, cell-type-specific microarray data (including their own) in order to ask the question: can they detect evidence of contamination, “stress effects” (more on that below), or any other kind of peculiar artifacts stemming from the purification (“fishing”) methods themselves? The short answer: Yes they can.

Some methods are fairly low throughput – fishing out one cell at a time.  The Manual cell sorting method (a home grown method) dissociates brain tissue, keeps the cells alive in artificial cerebrospinal fluid (almost literally seawater), and then the cell fisherman/woman hand picks labeled cells from the cell suspension with a glass pipette under a microscope (how they’re labeled isn’t terribly relevant to this discussion). This would be like collecting seawater, transferring the fish to less dense holding tanks with artificial seawater and then sorting the yellowfin tuna from the chub mackerel, etcetera. Another of the lower throughput methods is called Laser Capture Microdissection (LCM), where the extracted mouse brain is preserved through formalin fixation or flash freezing. Then thin tissue sections are made with a microtome, and individual cells are carved out of these tissue sections with a laser beam. This would be roughly approximate to freezing a volume of seawater, and then carving out the frozen fish of choice with a laser beam (sounds complicated). The primary difference between these two methods is that Manual sorts dissociated cells, whereas LCM extracts cells from intact, but preserved tissue.  Methods like fluorescence activated cell sorting (FACS) and immunopanning (PAN) also sort dissociated cells, and with the aid of flow cytometry, automated fluorometry, and/or the power of antibody selection (cell-type-specific bait), these methods greatly exceed the yields afforded by Manual cell sorting (imagine a dense network of narrow canals in which each fish is entrained in a high velocity stream, and an automated detection system diverts tuna into one channel, chub mackerel into another, and dolphin into another). Finally, a method called translating ribosome affinity purification (TRAP) bypasses the need to sort cells and “pulls down” tagged ribosomes, mRNAs in tow, from non-preserved tissue homogenate (a process which defies fishing analogy).

As you might expect, Manual cell sorting, along with FACS and PAN, achieve the highest purity (lowest amount of contamination), whereas LCM and TRAP show strong evidence of contamination from off-target cell types. Another concern is that the stress of dissociating cells or maintaining them in artificial media may perturb gene expression (think nervous, angry, wild fish in a cramped fish tank). However, only in the case of PAN data is there evidence of these effects (elevated levels of stress-response, cell death, and immediate early genes). Finally, the TRAP method extracts only mRNAs that are actively being translated, thus differences between TRAP data and data obtained by other methods may also reveal patterns of posttranscriptional regulation. For the full story, please refer to the paper.

addendum: see also Okaty et al. J.Neurosci. 31(19):6939-6943, 2011