Marc Juul BioFAB

I am from the BioFAB, which stands for the BioFAB International Open Faciltiy Advancing Biotechnology. So, the BioFAB is a 2-year grant from the NSF and we try to tackle this problem that exists in biology that parts are not designed to be composable. By parts I mean, promoters, binding sites, coding sequences, terminators, and by composable I mean, parts are not designed to work in the same way when you use the different combinations.

When we use it in another context, and say you have a strong promoter. But in a different context, it's not a strong promoter anymore. We tried to characterize this. We did 7 different promoters that we tested, and we tried it with rfp and gfp, and you can see that there's a lot of variance. The heat maps don't really look the same. They don't even look like they might be slightly the same. The graph on the right here shows that rfp vs. gfp, if they were composable you would expect them to all fall on the line.

Okay, so this is only in ecoli. It is definitely a problem, there are composability issues with these and coding sequences going on. Here's a worst case scenario. Assuming that you have two ribosome binding sites, or one 5-prime gene, and now you want to use it in a different gene. So these different genes have slightly different codons in the beginning (maybe), and these might result in secondary structures like hairpins, and this greatly effects the translation. So we see different levels of presence.

So what's the solution to this problem? Our solution is to standardize not just the parts but also the way that the parts are put together. So we are standardizing the junctions, and we are standardizing which parts are the minimum set of components. This is a specific system out of BioFAB: bi-cistronic 5' UTR:CDS junction, and standard +1 promoter libraries. Most of the interactions that ccause lack of modularity happen between the coding sequence and a 5-prime UTR. This is a different type of 5' UTR as usually used because it has two binding sites, and two coding sequences. It has a different, very short coding sequence that causes a different ribosome binding site. The first ribosome binding site is translated rapidly. The second ribosome binding site can then be mutated to be strongly or weakly binding depending on the gene of interest that you want.

The advantage of this system is that the first ribosome attaches and starts translation, and it unwinds the DNA such that the hairpin structures that might form, depending on your gene won't form, and then your gene expression should be more reliable.

So here's the system being tested. They were tested with different promoters. As you can see, these two heatmaps are a lot more similar than the first ones I showed. So here, we see why standard gene expression matters, and how this system improves predictibility. In the first scatterplot, we have a bunch of combinations, and in the second one we made fusion proteins with a bunch of different genes of interest. Since it's really only the first part of the coding sequence that makes secondary structures with the 5' UTR, this seemed like a reasonable way of doing it. The first graph on the left is the data from the all of the candidates or promoters that we might use normally, and without the standardized junctions. On the right we see after adding the bi-cistronic 5' UTR, and a bunch of different genes of interest in the first regions of coding sequences. The predictibility of this system is greatly improved.

Using this system, we can get 94% chance of predicting of where we are going to end up for fluoresence-wise, being within a factor of 2 of where we want to be. This is punpublished BIOFAB data under Dr. Vivek Mutalik. If you want to build complex systems with multiple genes that need to be expressed at different levels for your pathway, well, if you don't have a predictable system, then you will have to test a lot of combinations - the more complex or parts your system, the more time you will be spending tweaking or changing. The ad-hoc approaches will cost you a lot of time. This is our attempt to make a library of standard parts. We're launching a nice little GUI in a few weeks where you can go in and slide a slider for the level of expression that you want, and then you can go and test those options.

There are a few limitations. We have a bunch of different bi-cistronic designs. People complain that the results are very similar. We might have problems with similar sequences that might not be desirable. There's a new set being tested, but for now there's a set of 5' UTRs, promoters and terminators. You can go to the biofab website, click on these, and download this. Thank you.

No, two-fold is not the best system we could do. There is a trend in the evolutionary biological systems to be more robust, so it would be sensitive to small changes, so anything that could be tweaked.. but this gives us the ability to get very close or fairly close to it in one try or a couple of tries, and so you can still apply approaches, one or a few variations as people already do, if you want to get a very ... rightn ow the rpoblem isn't getting the target value, it's figuring out what to do with this. We're not very good at picking what we need, because we haven't had this capability before, because people when they designed pathways, they had no way of predicting of what they want. They don't know they want 75% of maximal expression of this gene and 33% of this gene, they only know an approximation. The next step in engineering biology is getting better at the design phase so that we can utilize these improvements.

Ellen: Have you done an experiment in house, where you take a known pathway and try to put this control pathway on it? No, we haven't done this, but there is a syystem in Endy's lab, but there was a paper recently about rewritable genetic one-bit memory that requires very fine control. And I believe they are trying to use this system there in that paper.

Nathan: Marc, do you see any utilization of this type of fab process, any potential security implications that come with regard to fabricating a product towards a specific function, and any associated customers or screening that might or could be involved in such a process? Could that be seen as subverting as subverting the sequencing screening in the federal guidance?

Marc: I think the interesting thing in the safety/security concerns is that we have a better ability, we need to do less work to get to the same work. So it's going to be more accessible and cheaper. I don't see any obvious way of integrating this into a regulatory system. I feel like the current regulatory system for self-checking is going to become less relevant and is probably going to become a lot easier both to design complex systems, even if we don't yet know what we are going to be able to do yet, we could imagine some bad stuff, but how feasible that will be is very unclear right now. It's hard to say exactly what will happen or when it will happen or what we can do about it.