Yeast knockout library allows for efficient testing of genomic mutations for cell-free protein synthesis.

Cell-free protein synthesis (CFPS) systems from crude lysates have benefitted from modifications to their enzyme composition. For example, functionally deleting enzymes in the source strain that are deleterious to CFPS can improve protein synthesis yields. However, making such modifications can take substantial time. As a proof-of-concept to accelerate prototyping capabilities, we assessed the feasibility of using the yeast knockout collection to identify negative effectors in a Saccharomyces cerevisiae CFPS platform. We analyzed extracts made from six deletion strains that targeted the single deletion of potentially negative effectors (e.g., nucleases). We found a statistically significant increase in luciferase yields upon loss of function of GCN3, PEP4, PPT1, NGL3, and XRN1 with a maximum increase of over 6-fold as compared to the wild type. Our work has implications for yeast CFPS and for rapidly prototyping strains to enable cell-free synthetic biology applications.

Introduction

Cell-free protein synthesis (CFPS), which was first used to decipher the genetic code, has recently made a renaissance with improved protein production capabilities and cost economics. CFPS has been particularly useful for proteins that are difficult to produce via standard in vivo methods (e.g., membrane proteins3, 4, 5 and proteins harboring non-standard amino acids6, 7, 8, 9, 10), high-throughput screening,11, 12, 13 and clinical manufacture of therapeutics.13, 14, 15, 16, 17, 18, 19 Additionally, the ease of protein production makes way for rapid prototyping in synthetic biology with applications in genetic circuits,20, 21, 22 metabolism, promoter libraries, and diagnostics, among others. This process uses crude cell lysates, or extracts, which contain the necessary translational machinery (including ribosomes, tRNAs, and aminoacyl tRNA synthetases) to drive protein synthesis in vitro. Substrates such as nucleoside triphosphates (NTPs) and amino acids are added to the extract along with salts and other factors to mimic the cellular environment. Once DNA encoding the protein of interest is added to the reaction, the product can be made in a matter of hours. Because the protein synthesis reaction now occurs outside the confines of the cell membrane, there are benefits including the ability to control the reaction components and conditions, as well as the ability to decouple cell growth from protein synthesis. Enabled by these benefits, several systems have emerged based on the lysates of Escherichia coli, Saccharomyces cerevisiae, wheat germ, insect cells, Leishmania tarentolae, Chinese hamster ovary cells, and tobacco BY-2 cells, among others.

In addition to containing the necessary elements for translation, crude extracts also contain many other enzymes that have the potential to positively or negatively affect protein synthesis. Examples of negative effectors could include enzymes responsible for degrading DNA and proteins, as well as using resources such as ATP that could be otherwise directed toward translation. Manipulations to the extract enzyme composition have shown utility in E. coli and wheat germ extracts. In one illustration, changes to extract processing methods removed protein synthesis inhibitors, such as thionins, and ribonucleases from wheat germ extract and allowed for production of up to 4 mg/mL dihydrofolate reductase in a continuous exchange reaction. Also, work in the E. coli system used genomic modifications to remove deleterious enzymes from cell-free reactions by deleting their corresponding genes from the source strain.9, 33, 34, 35, 36 These deletions included knocking out enzymes to impair amino acid and nucleic acid degradation pathways. For example, in order to address cysteine degradation, Calhoun and Swartz deleted the gene encoding for glutamate–cysteine ligase, which increased the lifetime of measurable concentrations of cysteine from 15 minutes to over 3 hours. Additionally, Michel-Reydellet et al. were able to stabilize linear DNA fragments by deleting endonuclease I. In a different genomically recoded chassis strain, Hong et al. observed a four-fold improvement in protein synthesis yields for products harboring non-standard amino acids through the deletion of five nucleases.

Recently, our lab has developed a novel CFPS platform in yeast that enables rapid protein expression from linear PCR templates.27, 37, 38, 39, 40 In terms of CFPS systems, our yeast platform benefits from being a microbe, a common protein production chassis, and a model organism. However, the platform currently suffers from low batch CFPS yields. Based on previous work, we hypothesize that this is due to the expected presence of nucleases and proteases, non-productive consumption of energy substrates such as ATP and other nucleotides, and low rates of translation initiation.27, 38, 39 We also note that unlike the E. coli system described above, yeast CFPS does not appear to suffer from amino acid substrate limitations. Guided by the results for wheat germ and E. coli CFPS systems above, we hypothesized that deleting potential negative effectors in the chromosome of the yeast crude lysate source strain could improve CFPS yields.

Many tools have been developed for engineering yeast. In particular, well established tools exist for the simple genomic modification of yeast cells through homologous recombination, and now also through the CRISPR system. As a model organism, the entire genome of yeast has been sequenced and all open reading frames have been characterized in a yeast knockout (YKO) collection. The use of this collection of strains can bypass the time investment for making a number of candidate mutations to characterize open reading frames. In a typical lab workflow, constructing single mutations in yeast takes approximately 7 days including primer design, PCR-based template construction, and knockout confirmation.

Here, our goal was to develop a method to rapidly test lysates from a series of single deletion strains in the YKO strain library, in order to efficiently identify gene deletions that can increase yeast CFPS yields. By leveraging the YKO collection, we aimed to reduce the total time for assessing a mutation by more than 50%. As a secondary objective, we wanted to assess the reproducibility of using the strains from the library along with our extract preparation methods. Thus, we set strict criteria for our work: the CFPS results came from two extracts of each mutant strain prepared from two separate fermentations.

We began by identifying possible negative effectors. Based on previous work primarily performed in yeast, but also E. coli, we targeted several relevant categories of enzymes. We chose knockouts of a protease (proteinase A, pep4Δ), two nucleases (poly-A specific exonuclease, ngl3Δ, and exoribonuclease, xrn1Δ), a phosphatase (protein phosphatase T, ppt1Δ), and two regulators of translation (eIF2 kinase, gcn2Δ, and eIF2B regulatory domain, gcn3Δ).9, 38, 45 Proteinase A (Pep4) is one of two proteases responsible for approximately 86% of all protein degradation in yeast. Next, given the benefit seen in E. coli CFPS upon the deletion of nucleases, as well as our reliance upon uncapped mRNA, we chose the strain deficient for exoribonuclease (Xrn1), a 5′-3′ exonuclease that acts on decapped mRNA. We also chose an exonuclease that acts in the 3′-5′ direction, Ngl3, which acts on polyA-RNAs. Protein phosphatase T (Ppt1) is a serine/threonine phosphatase. Based on previous work showing that our CFPS reactions are energy limited and that phosphate accumulates, we included a phosphatase mutant to explore the possibility that phosphatase activity in the extract non-productively cleaves high-energy phosphate compounds in our cell-free reactions. Finally, given that translation initiation is considered the rate-limiting step in protein synthesis, we chose the strains deleted for GCN2 and GCN3, which are inhibitors of translation initiation. Gcn2 phosphorylates translation initiation factor eIF2α. When phosphorylated, eIF2α inhibits eIF2B through interaction with the eIF2B subunit, Gcn3.48, 49

Strains harboring the above mutations were directly selected from the YKO library. As illustrated in Fig. 1, this library is a collection of yeast single mutant haploid strains, with each strain carrying one G418 resistance gene (KanMX) in place of every nonessential open reading frame of the S288c-derived BY4741. This is a commonly used laboratory strain that carries genetic auxotrophies (Supporting Information Table S1) to enable simple and fast genetic manipulation. We used the MAT a collection because the original source strain we used for yeast CFPS, MBS, is MAT a.

Fig. 1

A workflow diagram illustrates how the YKO collection is used for efficient screening of mutant strain extracts. Each nonessential ORF was deleted individually with the KanMX gene, shown here in orange, which imparts resistance to the drug G418. We selected a number of candidate strains from the yeast knockout collection that we subsequently grew in 1 L cultures, lysed, and processed into extract ready for use in cell-free protein synthesis.

Next, we grew 1 L cultures of each of the yeast mutants in duplicate in order to prepare extract and assess variability (Fig. 1). Fermentations were harvested at mid-exponential phase with an average OD of 11.50 ± 0.76. Representative growth curves can be seen in Fig. 2A and B. Previous work has shown that yeast harvested at approximately 12 OD is the most productive for CFPS.27, 39 OD measurements were taken over the course of the fermentation and the growth rate was analyzed during exponential growth. The gcn3Δ growth curve is offset from the others. This is due to a lower starting OD, resulting in the delay in exponential phase. All strains except gcn3Δ had comparable growth rates, as assessed by one-way analysis of variance (ANOVA) followed by Dunnett's test to compare each mutant to the wild type (p < 0.01).

Fig. 2

Growth of the strains used in this study. (A) Strains from the Saccharomyces genome deletion collection were grown in two to four 1 L cultures for subsequent lysis. (B) Growth rates were calculated using OD readings taken while cultures were dividing exponentially. Column values represent the mean with error bars corresponding to the absolute error of at least two independent growths.

After fermentation, the cells were processed into extract through several steps described in the Methods section. For each extract, a magnesium optimization was performed spanning 3–6 mM additional magnesium glutamate in the reaction (Supporting Information Fig. S2). The luciferase yield for the optimum magnesium concentration for each of the extracts is shown (Fig. 3). The data show the average of a total of eight CFPS reactions. The wild type BY4741 extract gave active luciferase yields of 2.98 ± 0.85 µg/mL, while the best mutant, xrn1Δ, showed yields approximately six times better with 20.00 ± 1.26 µg/mL. The gcn2Δ, gcn3Δ, pep4Δ, ppt1Δ, and ngl3Δ extracts yielded 3.95 ± 0.63 µg/mL, 11.16 ± 2.21 µg/mL, 15.23 ± 2.13 µg/mL, 11.32 ± 4.40 µg/mL, and 10.08 ± 1.75 µg/mL, respectively. All extracts except gcn2Δ showed higher yields that were statistically significant compared to wild type, as determined by one-way ANOVA followed by Dunnett's test to compare each mutant to the wild type (p < 0.0001). The gcn2Δ derived extracts gave yields that were not statistically different from the wild type. These statistical conclusions were consistent for the production of superfolder green fluorescent protein as well, as seen in the Supporting Information Fig. S3. The consistency of the sfGFP and luciferase results highlights the potential generality of our results.

Fig. 3

The impact of different extract source strains on CFPS yields. Extract from xrn1Δ, pep4Δ, and ppt1Δ strains was approximately 4- to 6-fold more productive than extract from wild type BY4741. Average CFPS yields for each strain were calculated from the yields of two separately grown and lysed extracts with four replicates performed of each reaction, for a total of eight independent reactions. The error bars represent the standard deviation of the replicates. Extracts that showed a statistically significant improvement over wild type (wt), based on ANOVA followed by Dunnett's test (p < 0.0001), are marked with an asterisk.

The CFPS results led us to several hypotheses about the composition of the lysate that will be the basis of future studies. We hypothesize that loss of XRN1 or NGL3 results in greater productivity in our extract due to a loss of nuclease activity. Xrn1 is a ribonuclease and could be active in extracts, reducing protein yield by destroying ribosomes. Ngl3, on the other hand, is a poly-A specific exonuclease involved in mRNA decay. We hypothesize that loss of Ngl3 in extract results in greater mRNA stability. Similarly, it is likely that loss of PEP4 increases extract productivity due to loss of protease activity. Pep4 is a vacuolar protease and could degrade the product as well as functional enzymes that generate our protein of interest. Also, it is possible that loss of PPT1 leads to a more productive extract by reducing the general phosphatase activity in the cell-free reaction. Interestingly, gcn2Δ did not show any significant improvement over wild type while gcn3Δ did. Loss of GCN3 is likely beneficial to extract productivity because loss of this subunit (eIF2α) removes translation initiation inhibition. In previous work, the deletion of GCN3 has been shown to remove translation regulation effects of phosphorylated eIF2, described above. Gcn2 is a known kinase that phosphorylates eIF2, upstream of the Gcn3 regulation.

In sum, our work demonstrates the merit of using the YKO library for rapidly assessing different single gene knockout strains on CFPS activity. Importantly, the time for making genomic mutations was eliminated in our study, allowing us to move directly to cell growth and extract preparation. Our data also highlight the robustness and reproducibility of our results. Looking forward, we propose that community resources, such as the YKO, should be leveraged in the field of cell-free synthetic biology to accelerate the ability to catalog positive and negative effectors on chassis strains.

Notably, the gene deletions that give beneficial improvements are capable of being ported into other strains of S. cerevisiae, such as MBS or S288c, which have produced higher yields of protein than that seen in BY4741, despite BY4741 being an S288c-based strain.38, 39 However, care will need to be taken to ensure that mutations beneficial to one source strain carry over to another. Additionally, it may be advantageous to explore whether beneficial mutations hold for a variety of extract processing conditions, such as lysis method. Nevertheless, once beneficial single knockout mutants have been identified, combinations of mutations have the potential for synergistic effects to improve yields even further. Looking forward, we hope to carefully explore multiplexing multiple knockouts in a future study, but this goes beyond the initial proof-of-concept study shown here.

Beyond improving protein expression yields, we anticipate that our general approach may be useful for studying metabolic pathways, genetic circuits, and promoter libraries in different strain backgrounds in a wide range of organisms. Not only do knockout libraries exist in other model organisms, such as E. coli,52, 53 but with the ease of the CRISPR/Cas9 systems54, 55 for eukaryotic genomic modifications, it is likely that more libraries will be prepared from commonly used eukaryotic cells, such as CHO cells.56, 57 We anticipate that combining cell-free systems with different strain collections will allow for a system-wide approach to gain a better understanding of how biological systems perform their versatile roles.

Methods

Cell growth and crude extract preparation

Yeast strains were from the YKO collection. Strains grew in 1 L cultures at 30 °C with shaking 250 RPM in 2.5 L Tunair full-baffle shake flasks (Sigma-Aldrich, St. Louis, MO) in SC media buffered to pH 5.5 and supplemented with 50 mM phosphate from an OD of 0.48 ± 0.12 to 11.50 ± 0.76. Methods for crude extract preparation from S. cerevisiae strains were identical to the methods described previously. Briefly, cells were lysed using high-pressure homogenization and dialyzed for buffer exchange.

Cell-free protein synthesis

CFPS reactions were carried out as described by Hodgman and Jewett, except that reactions were supplemented with 0.4 mM cAMP. The total protein concentration of the yeast extract was 8.90 ± 3.22 mg/mL as determined by Bradford Assay using commercially available reagents (Bio-Rad, Hercules, CA) with bovine serum albumin used as the protein standard. Cell-free reactions incubated for 5 hours at 21 °C, at which point the reaction had run to completion, and were then placed in ice and immediately assayed for active luciferase yield. The amount of active luciferase was determined by adding 2 µL of the CFPS reaction to 30 µL of ONE-Glo Luciferase Assay System (Promega, Madison, WI) in a white 96-well plate. Total luminescence was measured every 2 minutes over a 20-minute interval using a BioTek Synergy 2 plate reader (Winooski, VT) and the maximum output of relative light units (RLUs) was recorded for each sample. The values generated were then compared to a linear standard curve of recombinant luciferase (Promega, Madison, WI) added to the ONE-Glo Luciferase Assay System to calculate the active luciferase yield. The amount of active superfolder green fluorescent protein was determined as in Schoborg et al. All results are reported as means ± standard deviations. Statistical analysis was performed using JMP software (SAS, Cary, NC). A one-way analysis of variance was performed and followed by Dunnett's multiple-comparison test. A p-value less than 0.02 was denoted as statistically significant.

Competing interests

The authors declare no competing financial interest.