I have a second to blog - forgive me if it's full of typos (chances are, if you read this blog on a regular basis you're use to them) but it has been a while and I need to get back into the swing of things.
Last week, Jonathan Weissman came over to Harvard Medical School. I had the opportunity not only to hear him talk but also to attend a dinner with Dr Weissman and a handful of fellow postdocs. The Weissman lab has perfected a particular type of science, one that combines high throughput technologies and small reductionalist biology. This approach is the future of biomolecular science.
Let's take an example.
The Weissman lab has developed a strain of yeast where GFP is hooked up to a promoter that is activated by the unfolded protein response (UPR). Thus when the cell accumulates lots of unfolded proteins in the endoplasmic reticulum (ER), a transcription factor acts on the promoter and the cells turn green. The cells also express a separate red fluorescent protein under the control of a general promoter to control for gene expression, thus the ratio of green to red is a measure of UPR activation.Cells can be measured using microscopy or a FACS machine.
What they do next is perform these measurements in a collection of 400 different yeast strains. Each yeast strain is missing one non-essential gene whose product normally ends up in the ER. Thus for each gene you can obtain an indication of whether its translational product affects protein stability in the ER or whether if affects the ability of cells to detect unfolded proteins (i.e. the UPR machinery) - so far so good, in fact this at first glance seems to be another pedestrian type screen. Now here comes the trick. You cross each mutant strain to obtain double knockouts and test those. 400x400 = 160 000 strains = 160 000 measurements. With today's high throughput screening devices this is not an impossible task, but it's a lot of work. What do you get out of this exercise? Well superficially you find out what combination of genes result in an increase or decrease in unfolded proteins or in the function of the UPR machinery. BUT hold on a minute - what you have is something much more informative then that. Lets take gene #1. You now have a "profile" of gene #1 - you know what the UPR response is when it is missing in combination with genes #2, 3, 4 ... 400. Similarly you have profiles for gene #2 because you know the UPR response when it is missing in combination with genes #1, 3, 4 ... 400.
If two genes function together, then they should have a similar "profile". For example if gene #1 and gene #73 are two subunits of a stable complex, then when you eliminate #1 it should behave the same as a deletion in gene #73. Thus the profile for gene #1 (it's UPR state in combination with genes #2,3,4 ...400) should be similar to the profile of the other gene. Although in the latest version of the analysis the readout is UPR, you can use any quantitative measure. Previously in collaboration with several groups at the University of Toronto (Boone, Andrews, and Krogan who has since moved to UCSF), they performed a screen where colony size was the readout. They used that survey to map out proteins involved in secretion and identified components of the GET (Golgi to ER) complex. In more recent experiments the Hegde lab demonstrated that the mammalian orthologue to Get3, a cytoplasmic ATPase, is important for inserting tail anchored proteins into membranes. In fact last year I wrote about the Hegde paper, click here for a more detailed description of how tail anchored proteins are inserted by Get3/TRC40. If you abolish Get3 function, then the membrane fusion machinery is not properly localized and Golgi to ER traffic is blocked. Using the yeast knockout "profiles" where the quantitative measure is colony size Krogan Weissman and the gang discovered two proteins that co-clustered with Get3, these turned out to be membrane bound proteins that probably recruit Get3 to the surface of the ER. In more recent work the Weissman used the UPR "profiles" to identified other molecular players involved in this process (I won't say more as this data is not yet published).
But can these types of large scale analyses reveal anything else? Let me tell you the results are spectacular. Several members of the lab compared their small biology projects with some of the results from the Weissman screen and they were flabbergasted by the degree of agreement between our lab's biochemical data and the Weissman lab's high through put genetic data.
Other groups have launched into this type of profile based biology. Some commentators have seen these new large high throughput based scientific endeavors as the end of hypothesis driven research. However what the Weissman lab has done, is to use their results to drive their small biology research programs. They've looked at the results of their large scale analysis and used their knowledge of ER function to make guesses as to what might be going on between individual genes. They have substituted genetic screens with these high throughput quantitative screens. These are now referred to as epistatic miniarray profiles (E-MAPs). But just like genetic screens of old, they will only make any sense if they are combined with a more targeted approach that gives insight into the molecular machinery.
And studies using E-MAPs are starting to appear everywhere. Just this past month a study appeared in Science where the Krogan group used a similar aproach to identify a new component of the RITS complex, that would be the RNAi machinery that is responsible for silencing DNA in fission yeast (yes, I've written about this before, to check that entry out click here).
OK that's all the time I have to spare for today.
- Log in to post comments
Fascinating! Great post, holmes!