Getting cheap biofuel out of microbes involves careful tuning.
In the 1980s, advertisements for Holsten Pils claimed the pale lager was better because "most of the sugar turns to alcohol". Whether that did much for the flavour was a matter for debate, but for Amyris Biotechnologies it's pretty much the philosophy to getting fuel made by bugs onto the market.
"All the sugar has to turn to fuel," says Jay Keasling, a professor at the University of California at Berkeley and one of the founders of Amyris.
Although microbes are good at converting sugar into chemicals - as the beer business has demonstrated - they need more than a helping hand to get fuels into volume production. But, having already worked on a way to get a key malaria drug - artemisinin - on to the market in a much cheaper form, Keasling believes the optimisation is feasible.
For Amyris, the artemisinin project is something of a dry run for the bigger and much more lucrative market of biofuels. To license the technology needed to convince E coli and yeast to make the malaria treatment, the University of California demands that the manufactured drug is sold at cost - there is no room for profit. Amyris can use the technology and experience to work on boosting production beyond what a drug needs to what could drive a cheap biofuel that, in effect, recycles carbon dioxide into gasoline. However, engineering a microbe to turn out a chemical in even pharmaceutical quantities is no simple task.
"We started with E coli as our chassis because so much is known about this organism. It meant we could spend most of our time on engineering rather than concerning ourselves with biology," Keasling says.
However, unlike engineering in other fields, there are very few off-the-shelf components. For the past few decades, researchers have had to beg and borrow genes from their colleagues, and they don't always cooperate. One researcher saw Keasling's lab as a competitor and refused to part with one key gene, so the team had to search around for an alternative.
This replacement gene, found in plants, was able to synthesise chemicals called isoprenoids from a sugar source. Artemisinin, like the fuels that Amyris also plans to make, is an isoprenoid.
"But we got very low production of the isoprenoid from the native gene," Keasling complains.
One issue was that this was a plant gene, not one evolved by a microbe. One important aspect of DNA is that it is a redundant code. Within a gene, every three bases codes for an amino acid that will form the protein defined by the gene. But there are around 60 combinations of three-letter codes but just 20 or so amino acids in common usage. There is a lot of redundancy in the code.
For translation into protein to work, the cell has to recognise all 60 codons, making little chunks of RNA that attach to the relevant proteins, and which key into the complementary section of genetic code. Different cells favour different ratios of these RNA sequences. If you have a lot of codons that call for RNA that is comparatively rare in a cell, it will take a long time to assemble the protein. Alter the codons to favour more common sequences and you speed the whole process up. And that is what Keasling's lab did first. The result was a 100-fold increase in production rate.
That was only the start. It turned out that upstream processes would limit the output from the pathway that Keasling thought would be the best. They were planning to tap into processes and chemicals that were vital to the cell. Trying to increase production of artemisinin would stop the cell from growing and potentially kill it. On top of that, not all the genes needed for the pathway were known.
So, the team decided to bolt on a new production pathway that was better understood and used chemicals in the cell that were less vital to growth. This was the Mevalonate pathway. Linked to cholesterol, this pathway has had a lot of research effort lavished on it. "Cholesterol is a big business," Keasling notes. "And we could tap into it through acetyl-CoA. E coli produces so much acetyl-CoA that it has to break it down."
The breakdown process involves the production of an acidic chemical that lowers the pH around it. "E coli will literally pickle itself if you keep feeding it glucose," Keasling says.
Acetyl-CoA looked to be the ideal place to start and work started on inserting the genes that would take this feedstock and, through a series of enzyme reactions, convert it into artemisinic acid.
A big problem with the living cell is that everything depends on chemicals diffusing around its watery but crowded environment. The rate of that diffusion limits how quickly a chemical can be converted into its final product.
"It's like a plumber came in and threw all the pipes in the basement and expected the water to get from there to your shower," Keasling explains.
Nature has found a way round this: stick all the relevant enzymes on a scaffold so that they are physically close to each other. The intermediate chemicals can move from one enzyme to the next much more quickly than if they had to bump around the entire cell.
"The question was how many copies of each enzyme do we add to each scaffold? Do we add two, three, four or five copies?" asks Keasling. By engineering colonies of cells that would make different scaffolds, the team came up with a scaffold for three key enzymes in the ratio 1:2:2. The scaffold approach looks as though it could be an important feature of high-throughput biomanufacture.
"We now have some standard connections and we are trying this approach with other metabolic pathways. With the scaffold for artemisinin we got a 30- to 50-fold improvement in production. It was close enough to declare it a victory," Keasling claims.
But the work did not stop there. One enzyme was problematic in E coli: "It was the point at which you think, 'Darn, I've picked the wrong boat'," muses Keasling. The decision was taken to transplant the pathway to yeast.
The yeast approach had one big advantage, Keasling says: "We were surprised where the product ended up. Typically, you will look for the product everywhere." That means crushing the cells to find it, as well as trying to purify it from the water around the yeast.
"It turns out that artemisinic acid is toxic to yeast. The yeast turns on pumps that normally get rid of antibiotics. But the acid is pretty hydrophobic so that it sticks to the outside of the cell. The yeast invented its own purification process," Keasling explains.
The extraction process simply called for the cells to be centrifuged to concentrate them. A change to the pH caused the artemisinic acid to separate without killing the yeast - allowing it to be used again.
Gates foundation funds yeast approach
"At this point, we had two boats. The Gates Foundation gave us enough money to then have two teams compete to find the best. Both teams met their goals but, ultimately, we picked yeast."
However, for a product that's delivered in grams, scale is nowhere near as important as it is for fuel. The plan is to deliver 100,000kg of artemisinin a year. But that is only the equivalent of 800 barrels of oil. US president Barack Obama used a quarter of that amount of jet fuel to fly from Washington and back so he could deliver a speech in Iowa on Earth Day.
If equivalent yeast microbes were put to work making fuel, they would, in Keasling's estimate, wind up with gasoline at $400 a barrel. The target is to get that down to a few dollars: a difference of two orders of magnitude. But Keasling is confident that the target is achievable using current levels of understanding of biological systems. If the pils makers can do it, shouldn't bioengineers?