Michael Jewett sometimes finds himself at odds with biology.

As a bioengineer who uses synthetic biology to study ways to improve human and planetary health, Jewett needs microorganisms to manufacture specific compounds or follow the genetic instructions he’s given them. But the cells, evolved to ensure their own survival and reproduction, have other priorities. The microbes could, for instance, deactivate foreign DNA segments — or transgenes — inserted into their genome by Jewett. Or they might direct cellular resources towards their own needs instead of towards Jewett’s research goals.

How synthetic biologists are building better biofactories

So, he changed tack. “Rather than fight this tug of war that exists between what cells want to do and what we as engineers want to do, we kind of just cut the rope,” says Jewett, who works at Stanford University in California. “We rip off their cell walls and we collect the insides, and then we use the insides as a molecular factory to carry out the process of biology.”

It’s not a new idea: systems based on cell-free protein expression helped researchers to decipher the genetic code in the 1960s. But the technology is still evolving. “I would describe it as a very ‘in’ trend,” says biophysicist Vincent Noireaux, whose team at the University of Minnesota in Minneapolis developed its own system.

Although their composition varies, cell-free protein-expression systems generally contain all the machinery necessary to transcribe DNA into messenger RNA and translate that into proteins. These systems can be based on bacterial or eukaryotic cell lysates or assembled from purified components. The latter technology is called Protein Synthesis Using Recombinant Elements, or PURE.

With these methods, protein designers can screen hundreds or thousands of options without modifying and growing cells. This shortens the design–build–test–learn cycle from days or weeks to hours, says Wilson Wong, a synthetic biologist at Boston University in Massachusetts. Scientists can manufacture substances that would be toxic to living cells or get chewed up or altered by the cell’s enzymes. Some researchers are developing freeze-dried formulations to generate on-demand synthesis systems for biological medicines such as vaccines, or to serve as portable test kits for water-quality testing, point-of-care diagnostics and other purposes. Meanwhile, synthetic biologists are exploring PURE systems as a starting point for artificial cells.

“There are a lot of things going on,” says Noireaux: “a lot of new applications that are demonstrated, new cell-free gene-expression systems showing some new properties.”

Got lead?

Cell-free systems promise speed, customizability and convenience — but at a cost. The reagents needed to run a litre-scale cell-free system could cost more than US$4,000, and most studies report making less than two grams of protein, making these systems an order of magnitude more expensive than cell-based protein-synthesis ones. To keep high-throughput screening affordable, some researchers work at sub-microlitre scales.

And whereas bioengineers can culture a vat of bacteria or yeast synthesizing the desired products continuously by removing products and adding fresh medium, cell-free systems exhaust their energy supplies quickly. “Whatever you’re trying to do, it has to finish in a few hours,” says Wong.

That’s not a problem for cell-free sensors, which need only to detect a molecule of interest and produce a measurable signal in tens of minutes. “Nature developed all these sensors already,” says Jewett’s collaborator Julius Lucks, co-director of the Center for Synthetic Biology at Northwestern University in Evanston, Illinois. That’s because cells often need to recognize and respond to metals, antibiotics and other small molecules. Although scientists can use cells as living biosensors, the challenges are substantial. The cells must be kept alive and the molecules of interest must make it through the cell wall and membranes, Moreover, researchers run the risk of releasing genetically modified organisms into the environment.

Going cell-free meant that Lucks could build a minimal biological system without the hassle of using living organisms. In 2020, his group described a chemically defined system called ROSALIND that uses DNA-binding transcription factors as sensors. The transcription factors block RNA synthesis until they encounter a specific contaminant or molecule of interest¹. Once that happens, the protein falls off the DNA and the gene is transcribed, producing a signal. As a proof of principle, the team successfully tested freeze-dried ROSALIND systems, detecting copper and zinc in samples collected from municipal water-supply systems contaminated by a 2018 wildfire in California.

How cell-free processes could speed up vaccine development

But cell-free synthesis really shines when applied to speedy, iterative protein engineering, which can involve designs informed by machine learning. Lucks, Jewett and their colleagues used this approach to fine-tune another transcription-factor-based biosensor — this time, to detect lead². Lead is a tricky contaminant because many sensors that recognize it also stick to zinc, which is not generally considered harmful. To make a lead-selective sensor, the team fed data on a wide range of mutant proteins that could detect lead into a machine-learning model and used the results to design the next set of mutants, which were then tested in a cell-free system. Within a few rounds, the team had a sensor capable of detecting lead at as little as 5.7 parts per billion.

“I think that cell-free systems are going to be playing an increasingly important role in science because of the advantages they have in facilitating accelerated expression of tens of thousands to hundreds of thousands of proteins,” says Jewett.

Some of Lucks’ sensors are available from the non-profit reagent repository Addgene. His ROSALIND papers also include Microsoft Excel files with detailed instructions for the reactions. “We really try to make it as user-friendly as possible,” he says. “If they can assemble an in vitro reaction at all, they can set it up and run it.”

Protein-expression stews

To test the sensors, Jewett’s team grew the laboratory bacterium Escherichia coli, cracked the cells open and used their contents as a protein-expression stew. Compared with PURE systems, such ‘lysates’ often offer higher yields at lower costs, making them ideal for protein production, says Emily Chen, a production scientist at New England Biolabs (NEB) in Ipswich, Massachusetts. The biotechnology firm’s NEBExpress protein-synthesis system, for instance, is based on E. coli lysates.

Buying cell lysates can be a good deal, Chen says, especially for start-up companies and laboratories that don’t have all the equipment needed for cell-based protein production. Other commercial E. coli systems include Noireaux’s myTXTL, sold by Daicel Arbor Biosciences in Ann Arbor, Michigan; CFPS kits by Ginkgo Bioworks in Boston; and CFXpress by GenScript in Piscataway, New Jersey. Some providers also offer formulations that can improve the production of hard-to-express compounds. For example, Gingko’s CFPS Premium kit includes a solubility enhancer for proteins that are particularly large, insoluble or in need of extra folding assistance, says Cynthia Collins, general manager of Gingko’s reagents business unit.

Will self-driving ‘robot labs’ replace biologists? Paper sparks debate

Researchers who expect to burn through a lot of lysates can save money and customize the reagents by creating their own, but Noireaux suggests that scientists new to cell-free protein-expression systems get their feet wet with standardized, commercial products first. Lysates, whether home-brewed or commercial, must be supplemented with components such as glucose and nucleotides. These reagents help to replicate the cellular environment, provide energy and raw materials for protein translation and boost productivity, says Jewett. His lab found ways to cut costs by working out which ones are essential to the process. The researchers tested 1,231 formulations using lab-made lysates and optimized the recipe to just 12 reagents. This reduced costs by 95%, to less than $100 per gram of protein produced³.

Lysates made using other cell types are also available. These offer advantages such as the ability to complete proteins with post-translational modifications, including sugars or ubiquitin molecules, that off-the-shelf bacterial systems cannot. Historically, researchers often used lysates derived from immature red blood cells, or reticulocytes, obtained from rabbits with anaemia. But that model raises both ethical-animal-use and scientific concerns. Its translation process is unique, warns Evan Karousis, an independent researcher and lecturer in Bern. Unusually, reticulocytes don’t require an mRNA ‘cap’ to protect the mRNA and promote translation; they initiate translation at non-canonical sites; and they possess enzymes that can degrade RNA transcripts. Karousis had to stop a three-year project when he found out that rabbit reticulocyte lysates used a non-standard approach to the RNA decay process that he was investigating⁴.

Karousis suggests that researchers studying translation, rather than just looking to make proteins, find a lysate system that sticks closely to the biology they’re studying. To that end, the research group he previously led at the University of Bern developed techniques to make lysates from various human cell lines. They generated cell-free translation systems using cells derived from cervical cancer (HeLa), embryonic kidneys (HEK-293),and cancers known as neuroblastoma (SH-SY5Y) and osteosarcoma (U2OS)⁵. Of those, lysates from HeLa cells that adhered to a culture dish offered the most efficient translation, he says.

Plants, too, offer convenient cell-free systems — lysates from the cells’ light-harvesting organelles, called chloroplasts, are particularly useful, says Henrike Niederholtmeyer, a synthetic biologist at the Technical University of Munich in Straubing, Germany. That’s because plant bioengineers sometimes prefer to express transgenes in the chloroplast rather than in the nucleus. Nuclear-DNA modifications can cause transgenes to be silenced or their expression to be unpredictable. Biological containment is also a concern because nuclear DNA travels in pollen. Using engineered chloroplasts avoids those problems.

But chloroplast engineering isn’t easy, says Niederholtmeyer. And it’s slow: to test genes and proteins, scientists must wait for the plants to grow. So, it’s useful to test transgenes or regulatory sequences in chloroplast lysates first. She, Jewett and their colleagues compared chloroplast extracts from wheat (Triticum aestivum), spinach (Spinacia oleracea) and poplar (Populus × canescens), and found that they could use the spinach system to predict expression in the distantly related wheat⁶.

It takes a lot of leaves to obtain sufficient extract from just the chloroplasts. Niederholtmeyer says that her student bought a local supermarket’s whole spinach supply on a weekly basis, and they scaled their reactions down to 400-nanolitre volumes. But the study suggested that spinach-chloroplast lysates could be a reliable proxy for other plant species.