20.109(F09):Biofuel Production on Scaffolds
Kelly Drinkwater and Tina Tallon's research idea proposal project for 20.109 fall 2009. GO TEAM GREEN
Goal: use the modular protein scaffold model described by Dueber et al to improve bacterial production of biofuels.
Approach this problem from two perspectives: (1) get more biofuel; (2) try new methods of optimizing the protein scaffold.
(Alt. idea: can we make some use of a synthetic liposome/endosome/organelle to localize the reactions, per UCSF iGEM project?)
NB: We will be focusing on EtOH synthesis from pyruvate, not on digestion of biomass/feedstock to sugars (and eventually to pyruvate). That does need to be addressed though.
Global warming sucks. Biofuels rock. Ethanol is clean-burning and can be made from biomass (useless plant matter, basically). It would be awesome to make bacteria that ate biomass and made ethanol.
Artificial scaffolds can dramatically improve the yield of metabolic pathways, even those that already have a high output.
Our project is twofold. We want to explore new scaffold construction parameters to be optimized -- namely, trying different ligand/binding-domain pairs and changing the type and length of peptide linkers used to build the scaffold. We will use ethanol production as a test pathway because making ethanol is useful.
Many other projects have worked on getting microorganisms to produce ethanol, but the majority have relied on selection to obtain high yields. Our project is novel in the biofuels field because we're using a rational method, the scaffold, to increase yields.
Within 'theoretical' synthetic biology, our project is novel because we're exploring new ways to improve/optimize a scaffold system for a given synthesis pathway.
Choose biofuel end product
- Synthesis pathway is well characterized and fits into scaffold model (simple -- enzyme A, enzyme B, enzyme C...)
- Extractable from bacteria in some reasonably easy way
- Can be made from "common ingredients"
- Environmentally friendly to use
- Must be tolerant to high levels of chosen fuel
- Must be relatively safe to use
- Must have high baseline level of ethanol production
- Must be relatively easy to grow (can grow in relatively simple, easy-to-maintain environments)
How to optimize scaffold
- Try different linkers -- look up fusion-protein-ology
- Shape / stiffness -- can we adjust the orientation of the enzymes?
- Is there an easier way to optimize the stoichiometric ratios of the enzymes?
- Random combinatorial strategy a la Brainbow?
- Try different ligands / binding domains
- Reduce ligand tags' detrimental effect on enzyme efficiency
- Avoid ligand / binding domain pairs that crosstalk with our host cell
- Can we incorporate something that makes fuel extraction easier? Export fuel molecules from the cell maybe?
The Fuel Molecule
Pyruvate + pyruvate dehydrogenase -> acetaldehyde + alcohol dehydrogenase -> ethanol
- We should also try growing the experimental organism in different carbon sources to see what makes the most ethanol, because it's an easy manipulation, and because it will inform the next step (getting bacteria to digest biomass).
- Both acetaldehyde and ethanol are cytotoxic in huge quantities, but acetaldehyde is not cytotoxic in very small quantities. So we're kind of missing out on one of the big benefits of the scaffold method (avoid buildup of toxic intermediates). However, we are still getting the main benefit, which is increasing the local concentration of intermediates without increasing the global concentration. This could help us avoid having alcohol dehydrogenase run "backwards", removing ethanol from the system.
Assay for EtOH production
- Assay for ethanol concentration in media: Microfluidic
- Alternatively, use the Breathalyzer reaction. It doesn't sound like we could add AgNO3 and dichromate to an indicator medium and expect the cells to survive perfectly, but it'd be a great color-change assay. "Or we could just buy a Breathalyzer and spritz diluted culture onto it." This might or might not work, but it would be fast and cheap (and quantitative!).
- There's also gas chromatography
Note also: Head Space Gas Chromatography for acetaldehyde detection. This is nice but not strictly necessary for evaluating the efficiency of PDH which makes acetaldehyde from pyruvate.
If we can find an exporting pump/pore for the fuel molecule, then linking that into the scaffold should improve the rate of export. According to what Angela Belcher said about the Prather Lab's research, they're working under the assumption that ethanol exits the cells somehow, and apparently it's fairly easy to extract.
You can distill the ethanol -- it boils around 78°C -- and if you spin out your bacteria beforehand, then you can save them. (Although spinning out the bacteria from multiple gallons of culture might be a pain.)
There are a number of proprietary industrial processes that are less energy-intensive than distillation and are specifically designed to work well on liquid culture. If ethanol synthesis can be improved by using scaffolds, then the resulting high-yield strain could be sold or licensed to companies that have developed efficient extraction technologies.
Ethanol-tolerant E.coli strains or Zymomonas mobilis?
Zymomonas is ethanol-tolerant, and naturally better than E.coli at digesting hexoses (Chen et al). But there has been success in importing Zymomonas' digestion pathways into E.coli. It's an advantage if you can use an organism with a long history in the lab, because there's a much wider range of tools available for working with it.
E.coli! E.coli KO11 is a strain from the Ingram lab. They used repeated rounds of selection, alternately selecting for ethanol tolerance and ethanol production, to develop a strain LY01 that is both highly ethanol tolerant and has high production -- in fact it was better than either yeast or Zymomonas.
Yeast are also often used because they're naturally pretty good at fermentation (they make beer!). Why NOT yeast? They don't possess innate pathways to metabolize both pentoses and hexoses. Biomass is a hugely chemically diverse feedstock with all kinds of carbohydrates and other stuff in it, so you need a host organism with versatile digestion.
Improving The Scaffold
Ligand / binding domain pairs
One reason to try different ligands / binding domains is to find a ligand tag that doesn't affect the enzyme it's added to. Dueber et al showed that the addition of ligand tags does have some detrimental effect on the efficiency of pathway enzymes. For natively low-yield pathways whose output is drastically improved with scaffold (77 fold in the paper), this detrimental effect doesn't matter much. However, for pathways with a higher native yield, there may not be as much of a fold increase in output with scaffold, and the effect of adding ligand tags could be significant.
The other main reason to change ligands / binding domains is to find pairs that result in the most effective enzyme recruitment to the scaffold. Basically, this means you want the lowest Kd possible.
A third reason is to avoid binding domains that would bind to the wrong part of the enzyme+tag, rendering it inactive, or that would cross-talk with the host organism. Since we're using eukaryotic interaction domains in E.coli, that is not a huge concern, but it is a potential problem for using the scaffold system in eukaryotes. (Perhaps then you could use bacterial protein binding domains?)
From the Pawson Lab's protein interaction domain database, we selected those domains that are known to bind to a defined peptide sequence which could be added as a tag (rather than binding to a whole surface region which could not be made into a tag). We also excluded any domains whose cognate ligand had to be phosphorylated or otherwise post-translationally modified. Here are the ones we picked:
(See supplementary table 1 in Dueber et al)
What types of linkers exist for fusion proteins? Original scaffold paper used flexible Gly-Ser linkers. We don't know whether the linker type will have any effect, but it's worth testing this out both to improve ethanol yield in the current experiment and for future scaffold experiments -- linker type is a possible target for optimization, but we don't know if the scaffold effectiveness is sensitive to it. (I'm unsure how optimistic to be, because it seems like the most important mechanism of scaffolds is just that they increase the local concentration of everything -- it may not be possible to control the orientation of the pathway enzymes finely enough that that's even a consideration. However, I could easily see the length mattering, and as long as we're manipulating the length it's easy enough to try two types of linkers.)
See abstract of Arai paper:
- Flexible linker: (GGGGS)N, N = 3-4
- Helical (stiff) linker: (EAAAK)N, N=2-5
If one enzyme in a pathway is much faster than another, then it is probably beneficial for the complex to recruit multiple copies of the less efficient enzyme. To do this, simply include more copies of the binding domain that recruits that enzyme. This is totally modular, tunable to +/- one molecule of enzyme per complex -- a major benefit of the scaffold system. Because there are only two enzymes (PDH and ADH) in our synthesis pathway, we only need to make 1:n and n:1 ratios (assuming 2:2n is the same as 1:n, which assumption we may want to test).
- Figure out methods for EtOH assaying / extraction (and acetaldehyde extraction).
- Transform the ethanol production pathway into E.coli without scaffold, just to check that it works in the strain we're using, and as a baseline to compare to our experimental systems.
- Optimize pairing of enzymes with ligand tags -- see what has the least effect on enzyme efficiency. Try adding the ligand tag to either the N or C terminus of each enzyme.
- Make sure all of our proposed ligand/binding-domain pairs work at all: do pull-down assay and Westerns. Note which ligand/domain pairs pull down the most protein. There may be a trade-off between the ligand having little effect on the enzyme, and the binding domain having a lousy Kd.
- Choose linker. Build a 1:1 scaffold with a medium-length flexible linker, and one with a helical linker, and see which one is better.
- Optimize length of chosen linker. Build 1:1 scaffolds with linkers of varying length and see what's most effective.
- Optimize enzyme ratio. Build 1:n and n:1 scaffolds. (Also maybe build a 2:2n scaffold and see if it's comparable to the 1:n scaffold.) Vary n and choose the best ratio.
- Optional: once the ratio is pinned down, go back and switch up the linkers a little more to see if tweaking them a little more can increase the yield.
Note that this is a LOT of DNA construction (or synthesis), and that BioBricks is not suitable because it leaves scars. An alternative assembly standard may be in order.
Practicalities & Implications
"How many grad students does it take to screw in a biofuel synthesis pathway into a bacterium?"
People: 2 grad students, possibly one specializing in biochem and one in fusion proteins
Time: A year or so, optimistically.
Materials & Equipment:
- Stuff for bacterial culture
- E.coli strain KO11 or LY01 or something from the Ingram lab
- Ethanol assay: plate reader or chromatograph
- Optional acetaldehyde assay: chromatograph
- Materials for pull-down assay to evaluate recruitment of ligand-tagged enzymes by binding domains (check how this worked; need stuff for Westerns)
Money: "Enough grant to support 2 grad students"
What are the critical points of success/failure? If we succeed at point N and fail at point N+1, what have we contributed?
- Making baseline bacteria (no scaffold) that ferment glucose to ethanol via Pyr dehydrogenase / Al dehydrogenase
Other groups have already done almost exactly the same thing, though in different strains, etc. If we only got to this point it wouldn't really be a contribution.
- Find the ligand-tag/binding-domain pairs that work best for PDH and ADH (that is, have least detrimental effect on enzyme efficiency and recruit them to the scaffold most effectively)
We will have determined whether or not this is an important optimization to perform, and if so, we will have performed it on two enzymes of industrial importance. Incidentally, we will also have made a library of DNA constructs containing many different binding domains and ligand tags, which may come in handy for other scaffold users.
- Optimize linkers in scaffold protein
We will have determined whether or not this is an important optimization to perform, and presumably have figured out a reasonably efficient method for doing it without being driven crazy by DNA constructions. If it does turn out to be important, we will have established how best to couple PDH and ADH, and our result may also have relevance for scaffolds in general.
- Optimize stoichiometric ratio of PDH to ADH in complex
At this point we've basically completed our project goal. We'll have a viable, improved-yield system for producing ethanol (presumably from some combination of sugars). If its yield is competitive with systems produced by repeated selection, we have a winner.
Next Steps to Product
Connect our system to one for digesting plant biomass
There are three basic steps to produce ethanol from plant materials:
- Physical/chemical pre-treatment to obtain cellulose
- Chemical or enzymatic hydrolysis from cellulose to sugars
- Fermentation of sugars to alcohol
In this project we've addressed the fermentation step.
To make bacteria perform the hydrolysis step, have them secrete enzymes such as cellulase. The scaffold method is not particularly applicable here, because most hydrolysis reactions are one enzyme only.
It's possible that bacteria (or maybe fungi) could help with the pre-treatment step. For example, they could secrete proteases that break down wood constituents such as lignin. Biological pre-treatment could be a big improvement over chemical pretreatment if it yielded fewer toxic byproducts. This would most likely be a different strain or organism than that used for the fermentation step.
Perfect fuel extraction from culture
It would be nice if we could avoid having to kill all the cells or centrifuge gallons of culture medium, but since bacteria propagate themselves pretty fast, renewal of cultures is not a humongous problem. Alternatively, we could collaborate with another group or company that already has a good extraction procedure (several of these exist).
Avoid letting the system evolve out of the cells
Over many bacterial generations, the system may lose function. We should test this by repeatedly passaging a culture in the lab and assaying activity every N generations. If an ethanol factory operates for many years, it is to be expected that their 'working' cultures may have decreased ethanol yield over time.
We propose three methods for avoiding this loss. One, keep freezer stocks of the bacteria so that the working cultures can be periodically refreshed. Two, on a longer time scale, the DNA region containing the engineered system should occasionally be sequenced. Three, include a "kill switch" selection marker that will disappear if the ethanol production system is damaged or lost.
There are multiple possible kill switch configurations. Probably the simplest is to knock out some essential gene from the bacteria's genome, then include a copy of that gene with the plasmid that contains our system. However, we might also want to just integrate the whole system into the genome because that's generally more stable than plasmids. In that case, we could still knock out an essential gene and then put an extra copy of that gene inside the same operon as our engineered system -- that guards against loss of the entire system, but not fragments of the system.
Kill switches are also important for general biosafety, once we move from lab to industrial production. We may want to include one to make sure the engineered bacteria can't survive outside of our culture conditions -- perhaps make them dependent on a particular chemical with which we supplement our cultures.
Scale up and achieve positive energy balance
It's absolutely crucial to get more energy out than is put in -- without this, you don't have any kind of practical utility. Scaling up to giant vats of bacteria may or may not help us achieve a positive energy balance, depending if the usage of a given resource is constant or varies by production volume.
We can take advantage of known methods for growing E.coli on an industrial scale, since they're already used for other synthesis processes (insulin!).
What are possible use cases?
Giant vats of bacteria! Then ethanol is distributed via trucks (or, in urban areas, via pipeline), or used in-house to produce electricity, which may be more efficient. As stated, for factories we would adopt existing methods for using E.coli to synthesize gallons of stuff at a time.
Can we do distributed production (i.e. a vat of bacteria on every street corner) or does it only work in giant centralized factories? It depends on the ratio of a region's ethanol usage to its production of usable biomass or waste. This would increase the need for some pieces of physical infrastructure, like culture vats, but would decrease the need to transport stuff. (Then again, if the trucks are running on ethanol, it's clean burning and so transportation has a decreased carbon footprint!)
- Kitchen compost reactor -> stove fuel
You can cook a lot on an alcohol-burning stove. However, it's a little dubious whether a typical kitchen (or even a typical household, with paper waste and grass clippings) generates enough organic waste to metabolize into enough fuel to run the kitchen stove. Household-level production would also require a very simple, "kitchen grade" process for extracting the ethanol, and very stringent safety measures / kill switches in case people throw their culture down the sink. Overall, this is the least feasible of the several scale options.
(To say nothing about the typical household's willingness to keep a big tub of stinky bacteria in the garage! Maybe the banana/mint smelling bacteria are applicable here XD )
- Synthetic protein scaffolds provide modular control over metabolic flux. Dueber JE, Wu GC, Malmirchegini GR, Moon TS, Petzold CJ, Ullal AV, Prather KL, Keasling JD. Nat Biotechnol. 2009 Aug;27(8):753-9. PMID: 19648908 Link to full text PDF supplementary materials
- L. O. Ingram, P. F. Gomez, X. Lai, M. Moniruzzaman, B. E. Wood, L. P. Yomano, S. W. York. "Metabolic Engineering of Bacteria for Ethanol Production." Biotechnology and Bioengineering 58.2-3 (1998): 204-14.
- Perttu E.P. Koskinen, Steinar R. Beck, Johann Orlygsson, Jaakko A. Puhakka. "Ethanol and Hydrogen Production by Two Thermophilic, Anaerobic Bacteria Isolated From Icelandic Geothermal Areas." Biotechnology and Bioengineering 101.4 (2008): 679-90.
- Brent E. Wood AND L. 0. Ingram. "Ethanol Production from Cellobiose, Amorphous Cellulose, and Crystalline Cellulose by Recombinant Klebsiella oxytoca Containing Chromosomally Integrated Zymomonas mobilis Genes for Ethanol Production and Plasmids Expressing Thermostable Cellulase Genes from Clostridium thermocellum." Applied and Environmental Microbiology 58.7 (1992): 2103-2110.
- Page on Alcohol Dehydrogenase II describing the differences with ADHI -- I can't find a corresponding page for ADHI because I fail at search.