Yeast, the workhorse of synthetic biology, can convert sugars into chemicals, biofuels, and medicines. Metabolic engineers are constantly developing new tools to program and control S. cerevisiae, tweaking its genome and refactoring proteins to synthesize a milieu of compounds. But yeast – indeed, all life — is complex and intricate, and the specific genes that should be perturbed to increase the production of a compound are often tricky to delineate.
Now, in an experimental tour-de-force, researchers at Chalmers University of Technology in Gothenburg, Sweden have combined multiple tools – including computational models, in vivo biosensors, and CRISPR-mediated transcriptional activation and repression — to test many metabolic configurations at once, rapidly accelerating the pace of metabolic engineering in this model organism.
I sat down with Dr. Raphael Ferreira, lead author on this study, to learn more about the project, its challenges, and his plans for the future.
This interview on “Model-Assisted Fine-Tuning of Central Carbon Metabolism in Yeast through dCas9-Based Regulation”, published in ACS Synthetic Biology, has been edited for brevity.
Niko McCarty: Hi, Raphael. I’m excited to learn more about this study and, in particular, how it could accelerate metabolic engineering research. Could you walk me through the experiments and main findings?
Dr. Raphael Ferreira: In this study, we wanted to combine multiple methods to see if we could accelerate metabolic engineering with CRISPR screens. Our goal, or proof-of-concept, is simple: we want to identify guide RNAs that enhance the bioproduction of a chemical called malonyl-CoA. To do that, we first built a flux-balance analysis model for S. cerevisiae* to analyze which genes are associated with production of malonyl CoA. After building this model, we predicted that 168 genes could either be activated or repressed to enhance malonyl CoA production. Of course, in silico models have a lot of drawbacks — we make a lot of assumptions and these models don’t include everything that’s happening inside the cell.
After we identified these 168 genes, we transformed S. cerevisiae with a biosensor, previously developed by a postdoctoral fellow in our lab, that detects malonyl CoA and produces GFP (green fluorescent protein) in response. We then transformed these cells with 3194 different guide RNAs, which upregulate or downregulate these 168 genes, and sorted the cells based on GFP production to rapidly determine which guide RNAs led to enhanced titers of malonyl CoA.
At the end of all this work, we selected 49 guide RNAs that were enriched after the sorting, and these guide RNAs targeted a total of 46 genes.
(*Flux balance analysis, or FBA, is a mathematical method for simulating metabolism.)
Niko: Where did the inspiration for this project come on?
Raphael: A few years ago, there was an article published by George Church’s lab [Raphael refers to a 2015 Nature Methods paper by Kiani et al.] that showed that you could achieve both transcriptional activation and repression depending on where a guide RNA is targeted. Even though, now, I know that the effects are very modest, that paper is what gave me the inspiration to test guide RNA libraries in yeast to up- or downregulate genes. I wanted to see if I could adapt this and bring it to yeast. We decided to play with this idea by identifying genes relevant to a metabolic pathway of interest, such as malonyl CoA. Importantly, malonyl CoA is an industrially relevant compound that can be converted to 3-hydroxypropionate, a precursor for bioplastic.
To narrow down the number of genes that we had to test, it was logical to use a flux balance analysis model. Fortunately, we already had a biosensor for malonyl CoA production, which was created by a postdoc in the lab, and so everything kind of came together in a great way.
Niko: This project brought together so many experimental and computational tools – what in your background prepared you for this?
Raphael: I am French, and did my masters in Paris, focused on systems and synthetic biology, at The Center for Research and Interdisciplinarity (CRI). In March 2015, I moved to Sweden for my PhD at the University of Chalmers in Gothenburg under the supervision of Professor Jens Nielsen. He is a professor that is very well known in metabolic engineering, especially in yeast.
When I first joined the lab, I was assigned to work on lipid accumulation in yeast, but I preferred to focus instead on tool development in the context of metabolic engineering. I spent most of my time working with CRISPR, and really wanted to bring a variety of methods to yeast to accelerate metabolic engineering research.
As I moved through my PhD, I often found metabolic engineering research to be quite boring because, basically, many people just look at one pathway and then do knockouts or knock-ins. I was not very interested in this, so I really wanted to figure out a faster way to test many metabolic pathway designs at once. I’m not saying that our findings are amazing — this paper got rejected a couple of times, but at least we try to do something a bit a bit above just removing genes and seeing how the production of a chemical changes.
Niko: Well, you accomplished a lot during your PhD, something like nine first author papers. But I heard that this project took quite a long time – why is that?
Raphael: To be honest, every experiment in this paper took a while. The library sorting took a long time because we had to go to Denmark (Technical University of Denmark, or DTU) to do it, because we didn’t have a FACS machine in the lab. We also did a bunch of Illumina sequencing, which took quite a long time, and we decided to individually characterize a bunch of guide RNAs.
After sorting the cells and finding those with the highest GFP expression, we then individually validated that they were accumulating more malonyl CoA. In many cases, guide RNAs that had higher amounts of fluorescence, based on the biosensor, did not have as much malonyl CoA. It was a challenging project that required a lot of careful experiments and validation.
Niko: Are there any authors on this paper that contributed to specific aspects of the work?
Raphael: Well, the last three authors (Verena Siewers, Jens Nielsen and Florian David) are my co-supervisors. Two master students (Christos Skrekas and Alex Hedin) helped with the wet lab work, while Benjamín Sánchez built the computational model. Ben really added a unique aspect to the project, because I come from a very experimental background, and I think that the flux balance analysis model really brought a lot to this project.
Niko: What drew you to Harvard for your postdoctoral research?
Raphael: One of the papers that I published during my PhD related to genetic engineering using promiscuous guide RNAs, which are basically just guide RNAs that target repetitive sequences in the genome, and so can bind to many different sites. I developed a bioinformatic tool that could identify and predict promiscuous guide RNAs. Then, I visited MIT last winter for a visit and met a postdoctoral fellow, Oscar Castanon, from the Church lab. He told me about his project, which was basically using promiscuous guide RNAs to target repetitive sequences in the human genome, and we quickly struck up a collaboration. A lot of my papers are based on previous work from the Church lab, actually, so I knew that if I joined his group I would be able to jump on a project right away.
Niko: Thanks for talking with me, Raphael, and best of luck with your post-doc!
Biography: Dr. Raphael Ferreira is a postdoctoral fellow in George Church’s lab at Harvard. He previously completed his PhD at the University of Chalmers, and holds a Masters degree from the CRI in Paris. Dr. Ferreira has developed numerous tools for metabolic engineering in yeast, including the first proof-of-concept for guide RNA multiplexing from a single array in S. cerevisiae. His work has appeared in the Proceedings of the National Academy of Sciences, ACS Synthetic Biology, Microbial Cell Factories, and other journals.