Automated continuous evolution of proteins in vivo

Ziwei Zhong, Brandon G. Wong, Arjun Ravikumar, Garri A. Arzumanyan, Ahmad S. Khalil, Chang C. Liu

Posted on: 27 April 2020

Preprint posted on 22 February 2020

Article now published in ACS Synthetic Biology at

Evolution in a test tube: orthogonal replication plus automated feedback control allows for continuous evolution in live cells.

Selected by Pavithran Ravindran


Engineering of enzymes and other biological materials has been a central problem for bioengineering for over a dozen years. The advent of directed evolution has allowed for engineers to screen through an unprecedented number of variants and better understand the sequence to function relationship of these molecules1. However, screens to find which variants are capable of a desired outcome and multiple rounds of ex vivo mutagenesis limit such efforts. An evolution on the current methods would allow one to simultaneously vary a protein of interest in vivo and then screen for functionality. In this work, the authors decide to tackle exactly this problem and have invented a method they call ACE: automated continuous evolution.

Key Findings

In this preprint, the authors set out to develop a method to evolve proteins in live cells using the advantages of feedback control. In a beautiful collaboration, two groups that independently made advances on this problem came together to invent a technology that absorbs the best features of their respective systems. The Liu group recently created OrthoRep, which is a system for targeted, in vivo mutagenesis of a gene of interest2. Essentially, they use a error-prone, orthogonal DNA polymerase-plasmid pair to mutate genes of interest in yeast achieving a 100,000 fold increase in mutation rate to drive accelerated evolution. The Khalil group developed eVOLVER which is a continuous culture platform that allows one to control multiple parameters and apply selective pressures using feedback control and real-time monitoring of growth speeds3. Because this system is completely automated, the system does not need to rely on a priori selection schedules or algorithmic selection routines that can  drive the culture to an oscillatory growth regime or overshoots leading to culture extinction. In the ACE system, the authors set up eVOLVER to have two media inputs: (1) ‘no selection’ base media and (2) ‘full selection’ media that had maximum amount of drug or no nutrient. The system then would read out the OD of the culture and appropriately mix the two input medias to apply a constant pressure to drive the growth of the culture to a specific OD using a closed loop proportional-integral-derivative controller (Fig 1).


Figure 1: Schematic of ACE technology that combines the continuous mutagenesis of OrthoRep with the automated feedback control of eVOLVER. Figure taken from figure 1 of the preprint.


As a first test, the authors wanted to evolve Plasmodium falciparum dihydrofolate reductase (PfDHFR) to get drug resistance to pyrimethamine. To accomplish this, the authors encoded PfDHFR on OrthoRep in yeast that needed this protein for survival. They then found put these cells in eVOLVER that kept the growth of the cells at a target growth rate, thus maintaining constant strong selection pressure by the drug, causing rapid mutations to build up on the PfDHFR protein. After ~550 hours of ACE, five of the six replicates adapted to the highest possible concentration of the drug and the PfDHFR variants had converged upon known resistance mutations due to the high mutation rate of OrthoRep. The authors then chose to evolve a thermophilic Thermotoga maritime HisA enzyme to function in yeast at moderate temperatures. TmHisA is an enzyme important for histidine biosynthesis but does not complement the loss of the yeast ortholog HIS6, likely due to the thermophilic nature of the original host organism. To evolve this thermophilic protein into a mesophilic one they encoded TmHisA on OrthoRep in yeast that lack HIS6. With 600 hours of ACE, four independent replicates were able to grow in media completely lacking histidine meaning OrthoRep had once again mutated the TmHisA in a manner that allowed for the protein to function at a mesophillic condition. They found that the replicates had between 6 and 15 mutations to allow this protein to function under these conditions. This shows that ACE is durable enough to handle much more complex fitness landscapes and effectively navigate this to evolve proteins with interesting properties.

Why I chose this preprint

Directed evolution is an amazing tool that allows researchers to evolve proteins for specific novel properties. Sadly, the method for such a method typically involves multiple rounds of PCR to mutate the sequence of the protein, cloning to put the DNA into an expression vector, transformation into host organism of interest and finally screening. This leads to a time-intensive methodology that is limited. Here, the authors have developed a method to have completely hand-off evolution of proteins. An added benefit is that the enzymes are simultaneously screened for the ‘best’ variants in vivo for function. It will be very interesting to see how the authors and others are able to take this system to novel proteins for evolution and screen for more complex fitness landscapes and phenotypes.

Questions for the authors

  1. ACE looks like an amazing system for the evolution of proteins in yeast. How does this system compare to/is this different from other continuous evolution systems like PACE4?
  2. In this manuscript you have shown how well this system works for the evolution of proteins in yeast. With the interest in “smart-drugs”, what do you think the limitations are to moving this sort of system to mammalian cells?


  1. Arnold, F. H. Directed Evolution : Bringing New Chemistry to Life. Angew Chem Int Ed Engl. 57, 4143–4148 (2018).
  2. Ravikumar, A. et al. Scalable , Continuous Evolution of Genes at Mutation Rates above Genomic Error Thresholds Resource Scalable , Continuous Evolution of Genes at Mutation Rates above Genomic Error Thresholds. Cell 175, 1946–1957.e13 (2018).
  3. Wong, B. G., Mancuso, C. P., Kiriakov, S., Bashor, C. J. & Khalil, A. S. Precise , automated control of conditions for high- throughput growth of yeast and bacteria with eVOLVER. 36, (2018).
  4. Esvelt, K. M., Carlson, J. C. & Liu, D. R. A system for the continuous directed evolution of biomolecules. Nature 472, 499–503 (2011).




Read preprint (1 votes)

Author's response

Chang C. Liu shared

1.     ACE looks like an amazing system for the evolution of proteins in yeast. How does this system compare to/is this different from other continuous evolution systems like PACE4?
There’s room for many continuous evolution systems in the field and I’m delighted to see a flourishing of such systems in the past few years. In my view, the fundamental promise of continuous evolution systems is their ability to access scale and depth in evolutionary search. ACE and OrthoRep achieve scale and depth by turning rapid protein evolution into a simple process that requires just yeast culturing. Culturing microbes is easy to do in many independent experiments (scale) as well as for many generations (depth) so you can get long mutational pathways towards desired functions and do that many times over. PACE is also exceptionally good at achieving depth in evolutionary search, and pretty good at scale with recent improvements that simplify the original PACE “lagoon” design. A couple of important differences between ACE/OrthoRep and PACE do come to mind: OrthoRep is in a eukaryote whereas PACE uses bacteria, so some proteins and functions are more appropriate to be evolved using one or the other system; OrthoRep evolves proteins by coupling the desired function to the host yeast cell’s fitness whereas PACE evolves proteins by coupling the desired function to bacteriophage production, so different flavors of selections and functions are accessible; PACE uses the bacteriophage lifecycle for selection so generation times for evolution are very short, whereas ACE/OrthoRep uses yeast growth for selection so generation times for evolution are a few-fold longer, so PACE experiments can be faster than ACE/OrthoRep experiments (although for protein functions that require longer than the phage lifecycle to carry out, ACE/OrthoRep would be more appropriate). Etc.
2.     In this manuscript you have shown how well this system works for the evolution of proteins in yeast. With the interest in “smart-drugs”, what do you think the limitations are to moving this sort of system to mammalian cells?
There’s a lot to do in yeast! But you are right, it will also be important to go into mammalian cells especially to do things in the “smart-drug” arena like evolution of CARs and TCRs, possibly with the functional selection occurring directly in living animals. We are starting to work on this goal. Foreseeable challenges include making sure mammalian cells don’t reject a cytoplasmic DNA replication system (OrthoRep is cytoplasmic), and making sure the orthogonal plasmid of OrthoRep doesn’t productively integrate into the host genome, which is an additional risk in mammalian cells versus yeast because unlike yeast, mammalian cells undergo open mitosis.

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here

Also in the bioengineering category:

Scalable and efficient generation of mouse primordial germ cell-like cells

Xinbao Ding, Liangdao Li, Jingyi Gao, et al.

Selected by 05 March 2024

Carly Guiltinan

Cell Biology

Generalized Biomolecular Modeling and Design with RoseTTAFold All-Atom

Rohith Krishna, Jue Wang, Woody Ahern, et al.

Selected by 24 January 2024

Saanjbati Adhikari


Multi-pass, single-molecule nanopore reading of long protein strands with single-amino acid sensitivity

Keisuke Motone, Daphne Kontogiorgos-Heintz, Jasmine Wee, et al.

Selected by 04 December 2023

Benjamin Dominik Maier, Samantha Seah

Molecular Biology

Also in the synthetic biology category:

Enhancer cooperativity can compensate for loss of activity over large genomic distances

Henry Thomas, Songjie Feng, Marie Huber, et al.

Selected by 10 June 2024

Milan Antonovic


Discovery and Validation of Context-Dependent Synthetic Mammalian Promoters

Adam M. Zahm, William S. Owens, Samuel R. Himes, et al.

Selected by 21 June 2023

Jessica L. Teo

Synthetic Biology

Genetically encoded multimeric tags for intracellular protein localisation in cryo-EM

Herman KH Fung, Yuki Hayashi, Veijo T Salo, et al.

Selected by 16 January 2023

Martyna Kosno-Vega