Battling Malaria with ... Baker's Yeast?

According to the World Health Organization (WHO), malaria killed an estimated 655,000 people, mostly children, in 2010.  Artemisinin is an effective antimalarial  drug recommended by the WHO to be used in combination therapies. Artemisinin-based treatments could prove to be a silver bullet for the malaria scourge affecting developing areas of the world. There's just one catch: artemisinin is derived from Artemisia annua (Wormwood), an herb. Artemisia farming depends on the weather. Artemisinin may be only a small, small part of the overall plant mass, meaning that a great deal of resources (water, land, etc) are needed to produce small amounts of the desired drug. Thus, the current method of artemisinin production is unpredictable and inefficient. Queue, genetic engineering.

 Synthetic Biology Put to Work

Bakers Yeast (Saccharomycescerevisiae) is a common microorganism used in raising bread and fermenting alcoholic beverages. For obvious economic and gastronomic reasons, our society is good at growing yeast for cheap. Turns out, scientists have put a lot of effort into studying yeast genetics. Are you thinking what they thought at Berkeley? A fascinating article appeared in the Proceedings of the National Academy of Sciences in January 2012 [1] detailing the work of a group of Berkeley researchers who put two and two together. They manipulated the yeast metabolism to do a lot of the heavy lifting in artemisinin production. In plain English, they used their knowledge of the central dogma to engineer the yeast's DNA, RNA, and protein stages in a way that produced artemisinin. It was some bio-slick bioengineering.

Hurdles

Now, intuitively we might think "let's just cut all the genes necessary to make artemisinin from A. annua and paste them into the yeast genome. Yeast will make artemisinin, end of story." It's not quite that simple. Some key concerns of the research group were the following:

1. Former attempts at creating artemisinin in yeast used a very expensive sugar called "galactose". Could they switch to a cheaper input sugar?
2. Codon usage in the "transplanted" A. annua genes was not optimum for yeast translation (we'll talk about what that means). Could they optimize codon usage for yeast?
3. The metabolicpathway (molecular assembly line) in yeast was not balanced. It created too much of some pre-requisite molecules, and not enough of other molecules. The end result was sub-optimal amounts of end product. Could they balance the pathway and increase final output?

They addressed their concerns in the following ways, with the following results:

Could they switch to a cheaper input sugar?

Sugar is not a broad enough term in this case. What the Berkeley group was seeking was an alternative carbon source (fancy word for food) to feed to their yeast. Just like you and I enjoy the occasional ice cream, but also enjoy a hearty bowl of chips once in a while, yeast will thrive on many different carbon sources (including several sugars). While sugar serves as a source of carbon that can be used to build the artemisinin precursor molecules, ethanol is also a carbon source, though not a sugar. Even among sugars, not all are created equal. Different sugars have different effects on the yeast metabolism, which means that different sugars cause different genes to turn on or off. The process is similar to what my mom experienced when I was younger; if we wanted a special favor from her, we had to provide a carbon source to her in return. Raisins and cashews were pleasant gifts, but they just didn't have the same effect as chocolate in unlocking her generosity. Yeast is the same way. In the presence of glucose, a cheap sugar, a transcriptional repressor protein called Gal80p will bind to the promoter region of several genes essential to the production of artemisinin and prevent the transcription process from happening. With Gal80p in the way, RNA polymerase won't make mRNA copies of the genes, and no protein results. The missing enzymes aren't available to perform the chemical reactions in the artemisinin production line, and we're at a dead end. 

Galactose is to Gal80p as chocolate is to my mother; a key that unlocks. Galactose keeps Gal80p from repressing those key genes in the artemisinin pathway. Therefore, in order to produce artemisinin, galactose had to be present. But galactose is expensive! What's the solution? The Berkeley team found success by mixing a little bit of galactose with mostly glucose, which worked great. The galactose was enough to derepress the essential genes, and glucose served as primary carbon source. However, the team did not stop there. To further decrease the cost of production, and to simplify the process, they deleted the GAL80 gene altogether. By removing the GAL80 gene, there was no Gal80p repressor to unlock. Goodbye galactose. That's the equivalent of deleting mom's preference for chocolate, which would be a serious feat of bioengineering. 

As an interesting side note, the team also tested ethanol as a primary carbon source, and found that their yeast cultures produced nearly 8-times more artemisinin precursor than they did with glucose. The nice thing about deleting GAL80 was that they were free to use any carbon source, without the need to add galactose. Apparently the pathway from ethanol was more efficient in this case.

 

Could they optimize (find the perfect) codon usage for yeast?

Remember that a codon is a group of three DNA bases that code for an amino acid. While most organisms use the same genetic code (the same bases code for the same amino acids), there are multiple codons for each amino acid and sometimes an organism prefers one version over another. For example, CGT, CGC, CGA and CGG all code for Arginine. Most organisms will have tRNAs that match each of these codons. But, some tRNAs will be more abundant. Let's assume that in A. annua the tRNA for CGT is much more abundant that the tRNA for CGA. If a gene contains many arginines, all coded for using CGT, the protein will be produced more quickly and more abundantly that if all the arginines were coded for using the rare tRNA, CGA. Both versions of the codon will work, but the relative abundance of the tRNAs influences how efficiently protein is produced. It's similar to the supply of a textiles factory. By choosing to use imported cotton over local cotton, it is probable that new cotton shipments will take longer to arrive to the factory. If you run out of cotton, it will take a long time for a new shipment to arrive from India, and in the meantime the factory produces little or no textiles. A local supplier will probably be able to fill the factory's needs faster, meaning fewer interruptions in textile manufacturing. 

Proteins operate in a similar fashion. When the ribosome arrives at a particular codon, it effectively pauses and waits for the correct tRNA to snap into place. If there is no matching tRNA, the pause grows longer and longer. Eventually, it stalls and moves on to new, better mRNA strands, leaving behind an incomplete protein.

This issue of codon optimization is not good or bad, but merely another tool that can be used to control how much protein is expressed by a cell. To decrease the amount of finished protein, simply use suboptimal codons. In the case of the Berkeley team, the genes that they pasted into yeast originated in A. annua. Codon usage in A. annua is different than in yeast. They hoped that by replacing all the suboptimal codons with optimal ones, production of artemisinin precursor would increase.

Codon optimization of the genes in question was performed first by a computer. The Berkeley team had copies of the gene sequence (the sequence of bases that make up the gene) in text files. Previous molecular biology research had already determined which codons are optimal in yeast. Either by hand or using a computer program, someone on the team read through the sequence starting at the start codon, and checked each codon one-by-one. If a codon was suboptimal, they swapped it with an optimized version. Writing software to do this sort of thing is a classic example of bioinformatics, which will be covered in all its glory in another post.

With the new sequence in hand, they most likely sent the file to a company that specializes in DNA synthesis. Many companies exist that are able to synthesize DNA strands from electronic formats. They synthesize DNA with the desired sequence, and ship it back to the lab. The research team can plug the sequence into the yeast and continue on with their merry work. 

In the Berkeley team's case, they tested the newly optimized versions in yeast, and discovered that it really didn't make much of a difference in artemisinin production. Whatever bottlenecks were limiting production rate, suboptimal codons were not among them. 

Could they balance the pathway and increase final output?

Imagine a factory that produces chocolate peanut butter cups. Conveyor belt A carries empty chocolate cups to a machine that fills them with peanut butter. Conveyor belt B carries the finished product off to be boxed up. An overhead tube C feeds a constant supply of peanut butter to the machine. Suppose an impatient, hungry technician wants to go home early, and decides to speed up conveyor A. The natural result is that more empty cups will flood into the peanut butter machine, but they will continue to leave at the same rate they did before. The result is a backlog of empty cups at the mouth of the machine. They might even start piling up and falling off the conveyor belt onto the floor, which would be a terrible waste of good chocolate. So, the harried technician runs to the controls and increases the speed of conveyor B, but in his hurry, he overcompensates and now the machine, and conveyor B are both filling chocolate cups faster than conveyor A is supplying them. The result is that the machine is often goes through a filling cycle without any new cups, which sprays peanut butter all over the empty belt. Besides the peanut butter mess, since both A and B are running faster than peanut butter tube C, the reservoir of peanut butter in the machine quickly runs out and all that results when empty cups pass by is a sad sucking noise as peanut-scented air blows out of the nozzle. The technician ends ups staying late to clean up. 

The moral of the story is clear—don't make conveyor belt speedup decisions on an empty stomach. A second moral that in order to produce more, all stages of the process need to be ramped up. In designing yeast to produce artemisinin, simply up-regulating the genes that convert Acetyl-CoA to mevalonate will probably not end up producing more artemisinic acid in the end. It's likely that a backlog of mevalonate will result. The excess mevalonate will hang around the cell, serving no useful function, and possibly causing harm. Engineering a cell to produce a product requires some effort towards balancing inputs and outputs.

There are many possible ways of approaching this problem. Perhaps the enzymes that catalyze the reaction from Acetyl-CoA to Mevalonate are simply inefficient. Protein engineering is the field that specializes in optimizing proteins for a desired function (more on this in a future post). The Berkeley team took another tack. Rather than change the enzymes directly, they opted to alter the amount of enzyme present. More enzyme equals more Acetyl-CoA to Mevalonate (assuming there's enough Acetyl-CoA to feed the reaction). Less enzyme, less Mevalonate. 

The artemisinin-producing metabolic pathway (reproduced from [1])

To accomplish this, we return to the now familiar promoter region upstream of a gene. Remember that the promoter attracts RNA polymerase, which goes on to make mRNA. Some promoters attract better than others. Some promoters are repressed by specific proteins (like the Gal80p example), while others are activated by specific proteins. Some promoters have both activators and repressors. Promoters also work differently in different organisms. A reporter that works in E. coli might not work at all in yeast, or might work too well. Too well means that a cell allocates too many resources to producing whatever follows the super-promoter, and winds up sick or dead. I've personally made E. coli cells sick because they produced too much GFP. There are entire libraries of repressors fine-tuned to suit the needs of a biotech researcher.

The Berkeley team noticed that many genes in their artemisinin pathway were controlled by completely different promoters, which resulted in different enzyme levels at each step in the process. This is analogous to having different conveyor belt speeds in the chocolate factory. One precursor might be backlogged while another precursor is in short supply, all of which results in an inefficient process. To solve this problem, the Berkeley team swapped out several promoters in the mevalonate pathway and replaced them with matching promoters from the rest of the system. Remember from earlier that the team deleted the GAL80 gene so that galactose would no longer be necessary to derepress (unlock) transcription of needed genes. That's because Gal80p binds to the promoter region of the GAL1 gene. Another way to understand this is that many of the essential genes in the pathway were controlled by the GAL1 promoter. Understanding this, the Berkeley team replaced other promoters with the GAL1 promoter so that the entire pathway would be operating at the same pace. Using the same promoter for everything is like setting all the conveyor belts to the same speed. The enzymes will still catalyze reactions with different efficiencies, resulting in backlogs, but cells are resilient. At least by using the same promoter for everything, production levels are in the same ball park. 

Final Results 

The resulting strain of yeast from the Berkeley team's work was able to produce reasonable amounts of amorpha-4,11-diene from ethanol or glucose, which can then be converted to artemisinin in a laboratory. The paper explained difficulties with the process that have yet to be overcome, and no doubt it will take years for cheaper, yeast-produce antimalarial drugs to hit the streets. But, are you beginning to catch a glimpse of the power of biotechnology to solve problems? Originally it was necessary to wait months to grow fickle plants, having to worry about supplying water, fertilizing, planting, protecting the fields from hungry wildlife, harvesting,  and then extracting the final chemical product. Now it is possible, through similar fermenting technology used to make beer, to produce artemisinin. It can easily be scaled-up to make large batches. It's fast. It does not depend on weather or wildlife. No fertilizer needed - just sugar and some cheap broth.

By producing artemisinin with yeast, the drug can potentially be produced in greater quantities, more reliably—all for less money—which means it can reach more of those suffering from malaria. That's the problem-solving, bio-awesome power of biotechnology. The field is still in its infancy, and huge breakthroughs are on the horizon—breakthroughs that you can be part of. 

 [1] Westfall, PJ et al. (17 January 2012) Production of amorphadiene in yeast, and its conversion to dihydroartemisinic acid, precursor to the antimalarial agent artemisinin. PNAS, vol. 109, no. 3, E111–E118.

More on Synthetic Biology in General

 

---