A highly efficient human cell-free translation system

Cell-free protein synthesis (CFPS) systems enable easy in vitro expression of proteins with many scientific, industrial, and therapeutic applications. Here we present an optimized, highly efficient human cell-free translation system that bypasses many limitations of currently used in vitro systems. This CFPS system is based on extracts from human HEK293T cells engineered to endogenously express GADD34 and K3L proteins, which suppress phosphorylation of translation initiation factor eIF2α. Overexpression of GADD34 and K3L proteins in human cells significantly simplifies cell lysate preparation. The new CFPS system improves the translation of 5’ cap-dependent mRNAs as well as those that use internal ribosome entry site (IRES) mediated translation initiation. We find that expression of the GADD34 and K3L accessory proteins before cell lysis maintains low levels of phosphorylation of eIF2α in the extracts. During in vitro translation reactions, eIF2α phosphorylation increases moderately in a GCN2-dependent fashion that can be inhibited by GCN2 kinase inhibitors. We also find evidence for activation of regulatory pathways related to eukaryotic elongation factor 2 (eEF2) phosphorylation and ribosome quality control in the extracts. This new CFPS system should be useful for exploring human translation mechanisms in more physiological conditions outside the cell.


Introduction
Cell-free protein synthesis (CFPS) systems have become increasingly important to advance biotechnological and fundamental research (1). In CFPS systems that use cell-free extracts, an mRNA of interest can be translated in conditions that recapitulate protein biosynthesis in vivo and reveal insights into the translation process. Cell extract-based translation systems can also be used to overcome the inherent limitations of cell-based experiments by removing cellular membrane and natural homeostasis mechanisms that prevent the screening of many translational parameters.
Whereas cellular growth requires defined conditions, cellular extracts may tolerate substantial genetic or proteomic manipulation and unnatural or toxic components (2). In addition to functional and structural studies (3), proteins expressed in cell-free extracts can be used for selective and sitespecific labeling, stabilization of membrane proteins in a soluble state, and for optimizing the production of toxic proteins (4). In vitro translation systems also allow high-throughput screening of thousands of individual proteins translated from mRNA libraries (5,6).
To enable high levels of protein synthesis, CFPS systems require supplementation with exogenous amino acids, cofactors (proteins, small molecules, and inorganic ions), and energy resources (1), in addition to the translation templates provided in the form of mRNA, or DNA in the case of transcription-translation coupled systems. Many organisms have been used successfully as the source for extracts to prepare in vitro translation systems. Escherichia coli and wheat germ extracts have the highest protein production yields among the cell-free translation systems and are widely used to produce recombinant proteins (1). However, these systems often do not recapitulate conditions needed to translate mammalian proteins of interest.
To date, mammalian extracts have been limited to human cell lines and rabbit reticulocyte lysates (RRL). While these mammalian systems may be an adequate model for fundamental translation studies, and can provide a native environment for protein folding and post-translational modification, several shortcomings limit their functionality (1). Preparing RRL extracts requires labor-intensive maintenance, treatment, and sacrifice of animals, while commercially available lysates are expensive (7). Using rabbits as a source also limits the possibilities for genetic manipulations, including the ability to knock out or enrich specific proteins. Furthermore, RRL derives from a highly differentiated and specialized animal tissue, which limits the scope of its translation regulation mechanisms. Finally, RRL endogenously contains a very high concentration of globin mRNA, which outcompetes translation of exogenously added templates unless the RRL is pretreated with a nuclease (7). By contrast, the use of human cultured cells provides more flexibility, by allowing for precise genetic manipulation and the use of many different cell types (8)(9)(10). However, both RRL and human cell line-derived extracts have relatively low protein yields, which limits their functionality. Therefore, highly efficient human-based translation systems are needed to overcome the limitations of presently used mammalian CFPS systems.
A common limitation of presently-available mammalian translation extracts is the attenuation of translation initiation due to the phosphorylation of translation initiation factor eIF2 on subunit eIF2α. During translation initiation, eIF2 delivers initiator tRNA to the 40S ribosomal subunit in a GTP-dependent manner. After mRNA start codon recognition, eIF2-GDP is released from the ribosome and is subsequently converted to eIF2-GTP by the guanine nucleotide exchange factor eIF2B. The phosphorylation of eIF2α, which generally occurs in cells during stress (11)(12)(13), increases the affinity of eIF2 for eIF2B by nearly 100-fold, resulting in the sequestration of eIF2 from the translating pool and inhibition of translation (14). In mammalian cells, four known eIF2αspecific kinases can phosphorylate serine 51 of eIF2α (Fig. 1A). The PKR kinase is activated by double-stranded RNA, as found during viral infection or during in vitro transcription (15)(16)(17)(18).
During stress, eIF2α may also be phosphorylated by PKR-like endoplasmic reticulum kinase (PERK), general control non-derepressible-2 (GCN2) kinase, and heme-regulated HRI kinase (14) ( Fig. 1A). Phosphorylation of eIF2α can be bypassed in human cell extracts by the addition of the human GADD34 protein, an eIF2α-specific adapter for PP1 phosphatase, and/or by vaccinia virus K3L protein, a substrate for eIF2α-specific kinases that can act as a decoy (19)(20)(21)(22) (Fig. 1A). While eIF2α phosphorylation is thought to be the main limiting factor in CFPS systems, attenuation of the activity of other translation factors has not been studied in depth.
We describe here a highly efficient cell-free translation system based on genetically modified human HEK293T cell extracts. We find that overexpression of a truncated version of GADD34 and K3L in HEK293T cells efficiently reduces the phosphorylation of eIF2α and improves the translation activity of the resulting cellular extract, including robust formation of polysomes on exogenous mRNAs. This system can be used for 5' m 7 G-capped and IREScontaining mRNA templates, in mRNA-dependent or transcription-translation coupled reactions.
We also probed the regulation of eIF2α phosphorylation, the phosphorylation status of eukaryotic elongation factor 2 (eEF2) in these extracts and identify avenues for future optimization of the CFPS system that could enable reconstitution of translation regulatory pathways for biochemical and structural studies.

Results
Endogenous expression of GADD34Δ and K3L increases the translational activity of HEK293T cell extracts.
To develop a quick and robust method to generate translationally active human extracts, we first tested the hypothesis that endogenously expressed proteins GADD34 and K3L might increase the synthetic activity of the in vitro translation system. We deleted the N-terminal 240 amino acids of human GADD34, hereafter denoted GADD34Δ, as this deletion allows high levels of GADD34 expression without compromising its enzymatic activity (23). By employing the Sleeping Beauty transposon stable integration system (24), we engineered the HEK293T human cell line to express both the GADD34Δ and K3L proteins under the control of a doxycyclineinducible promoter (Fig. 1B) (See Materials and Methods). Stable integration of the expression construct minimizes expression variation between different lysates preparations. In addition, a tightly controlled inducible promoter bypasses potential toxicities due to the overexpression of these proteins. After optimization of doxycycline levels, GADD34Δ expression was detected in HEK293T cells without visible toxic effects on the cell's growth (Fig. 1C).
Efficient in vitro polypeptide synthesis requires substantial energy resources (3,5). To mimic the physiological environment and bypass the potential stringency of nucleotide triphosphates in the CFPS reactions, we supplied the human cellular extract with an energyrecycling system (2). Although the primary energy consumption of CFPS systems is aminoacyladenylate formation (25) and, therefore, the transformation of ATP to AMP and two inorganic phosphate molecules, the creatine kinase commonly used in energy regeneration systems synthesizes ATP from ADP and creatine phosphate (26). To overcome this mismatch, we added rabbit myokinase, which converts the AMP released after tRNA aminoacylation to ADP, the substrate of creatine kinase, by transferring the γ-phosphate from ATP to AMP (27). Furthermore, since several human translation factors use GTP (28), we added nucleotide diphosphate kinase Aleksashin, N.A. et al.

Manuscript
Page 6 of 30 which can maintain the steady-state level of GTP by transferring the γ-phosphate from ATP to GDP (29). Altogether, these three enzymes restore the concentration of ATP and GTP, which is required for efficient cell-free translation. The engineered HEK293T cell extract with GADD34Δ and K3L (hereafter called "engineered cell extract") was supplemented with a nanoluciferase (nLuc) mRNA, and the in vitro translation was carried out under conditions described in the Materials and Methods (8). The synthetic activity of the CFPS system was monitored by the accumulation of enzymatically active nLuc (30) and was found to be ~50-fold more active than extracts from HEK293T cells lacking GADD34Δ and K3L overexpression (hereafter WT extracts) based on the translation of enzymatically active nLuc (Fig. 1D). Moreover, the new translation system remains synthetically active for much longer periods of time, possibly suggesting the high stability of the mRNA in the CFPS system, and the increased activity of translation factors due to the modified energy regeneration system (Fig. 1D). In the above reactions, the reaction conditions were optimized separately in order to maximize the efficiency of in vitro protein synthesis in each extract.
To further characterize the activity of the optimized system, we measured the ability of the Next, we tested if the CFPS system based on the engineered cell extract can improve translation of mRNAs with different 5' UTRs. While the EMCV IRES containing mRNA helps drive strong protein expression without a 5' m 7 G cap structure (Fig. 1D), its large size and complicated RNA secondary structure may limit its applications. We tested the CFPS system with nLuc mRNAs using 5'-UTRs containing the cap-independent HCV IRES (31, 32), CrPV IRES (33,34), omega leader (35), or synthetic SIST sequence (36). However, we could not identify optimized temperature and ionic conditions for any of these 5'-UTRs that improved their Aleksashin, N.A. et al.

Manuscript
Page 8 of 30 translation to comparable levels as the EMCV IRES (Fig. 1G). The widely used capped human βglobin (HBB) 5'-UTR (37) shows substantial translational activity in the optimized CFPS system ( Fig. 1G), but not as high as with the EMCV IRES.
The activity of engineered HEK293T cell extracts is comparable to a HeLa-based in vitro translation system.
The expression and purification of recombinant GADD34Δ and K3L proteins add labor and cost to preparing CFPS systems (20,23). Therefore, engineering human cell lines to express these recombinant proteins should reduce the cost and time required to prepare extracts for CFPS.
Consequently, we compared the translational activity of a commercially available CFPS system with cell-free translation systems prepared in-house. The commercially available system based on the S3 HeLa cell extract contains ectopically purified GADD34Δ and K3L accessory proteins ( Fig.   2A). The homemade reaction was based on the ECE and contained all the required supplements (see Methods). The two translation systems were set up in parallel and in the same volume to test their activity, using nLuc or GFP activity assays to monitor the translation of EMCV IREScontaining and polyadenylated mRNAs encoding these proteins. When compared to engineered cell extracts prepared as described above, the translational activity of the commercially available HeLa translation system is not substantially higher than the in-house prepared CFPS system prepared from the engineered HEK293T cells (Fig. 2B, C). Thus the overexpression of the GADD34Δ and K3L accessory proteins in human cell lines rather than subsequent addition to cell extracts effectively promotes CFPS activity to the same extent while significantly reducing preparation time and cost (20,23).

Figure 2. Endogenously expressed GADD34Δ and K3L increases translational activity comparable to the addition of exogenously expressed accessory proteins.
A, Western blot showing the amount of the GADD34Δ expressed in the engineered HEK293T cells and supplemented in the HeLa-based commercial translation system. The asterisk indicates a nonspecific band in the HeLa extract. The gel is representative of two independent experiments. B, A time course of nanoluciferase synthesis in the CFPS systems prepared based on the engineered HEK293T cell extract and HeLa-based extract with recombinant GADD34Δ and K3L supplement. All error bars represent one standard deviation of three independent replicates. C, Cell-free synthesis of GFP in the two translation systems. Orange bars represent the HeLa-based extract with exogenous GADD34Δ and K3L added, while the white bars represent the engineered HEK293T cell extract with endogenously expressed GADD34Δ and K3L proteins. All error bars represent one standard deviation of three independent replicates.
Cellular overexpression of the GADD34Δ and K3L proteins protects eIF2α from phosphorylation.
The increased translational activity of the CFPS system based on the engineered cell extract seen above is consistent with the role of GADD34Δ and K3L in preventing phosphorylation of eIF2α in the in vitro translation reactions. To test this hypothesis, we probed the phosphorylation state of eIF2α in CFPS systems based on cell lysates from the engineered and WT HEK293T cells.
By employing phos-tag gels (38,39), which separate the phosphorylated and non-phosphorylated forms of eIF2α during gel electrophoresis, western blots of eIF2α revealed the baseline level of phosphorylated eIF2α in translation systems from WT and engineered cell extracts was similarly low after one hour of incubation, even though endogenous mRNAs were not removed (Fig. 3A).
However, addition of exogenous mRNA to CFPS reactions significantly increased the amount of the phosphorylated eIF2α in WT extracts (up to 28%) compared to the small increase seen in engineered cell extracts containing GADD34Δ and K3L (~5%) (Fig. 3A). This increase in WT extracts was independent of whether the 5' UTR of the mRNA was cap-dependent (HBB 5'-UTR) or used the long and highly structured EMCV IRES (Fig. 3A). Remarkably, the level of phosphorylated eIF2α was highest in the translationally efficient system based on the HeLa extract supplemented with recombinant GADD34Δ and K3L proteins (~44% and ~51% phosphorylation without and with exogenous mRNA added, respectively) ( 3A and B), even though these proteins are added at roughly the same concentration to the two CFPS systems ( Fig. 2A).
Next, we asked if the residual eIF2α phosphorylation in the CFPS system based on the engineered cell extract was mediated by a particular eIF2α-specific kinase and whether inhibition of this kinase could affect the protein synthesis activity of the system. Although the four known eIF2α-specific kinases are each activated by specific stress response signals, these signaling pathways might be dysregulated in the context of cellular extracts. For example, GCN2 kinase might be activated by the supplementation of the system with uncharged bovine total tRNA (See Materials and Methods). To test this idea, we checked the phosphorylation state of eIF2α in CFPS reactions pretreated with different eIF2α-specific kinase inhibitors (Fig. 3D). We used GCN2 kinase inhibitor A-92 (IC50 = 300 nM, (40)), PKR-specific inhibitor C-16 (IC50 = 210 nM, (41)) and PERK kinase inhibitor GSK26006414 (IC50 = 0.4 nM, (42)). Only the GCN2-specific kinase inhibitor prevented eIF2α phosphorylation at a concentration near its IC50 value, whereas the PERK and PKR-specific kinase inhibitors failed to prevent eIF2α phosphorylation, even at concentrations far above their IC50 values (Fig. 3D). While this result revealed the likely cause of eIF2α phosphorylation in the cell-free protein expression reactions , the decrease in eIF2α phosphorylation did not increase the translational activity of the engineered cell extract-based human translation system (Fig. 3D), indicating that eIF2α phosphorylation is not limiting translation in the new CFPS system. Although in the ECE-based CFPS system, eIF2α phosphorylation levels do not exceed a few percent (Fig. 3A),  Several factors may limit production of an even more active CFPS system by affecting translation elongation, including induction of the phosphorylation of eEF2 and activation of ribosomeassociated quality control factors.

Discussion
Here we describe the preparation of a highly-active CFPS system derived from human cells engineered to express GADD34Δ and K3L, which minimize eIF2α phosphorylation in the extract. We also implement a new energy regeneration system informed by the nucleotide pools that result from the different steps in translation. One of the most significant advantages of the translational system described here is that all components of the translational machinery are derived from human sources. While previous results in Chinese hamster ovary (CHO) cells that transiently overexpressed GADD34Δ also showed increases in the translational activity in a CFPS system (48), the use of CHO cells may introduce confounding variables to the study of aspects of human translation that are not conserved. Stable integration of the recombinant genes for GADD34Δ and K3L in the genome as described here should also enable more controlled preparation of cell extracts without complications due to transient transfection on cell viability (48). In the case of human derived CFPS systems, we find that HeLa translation systems can achieve high levels of translation (Fig. 2). Although this in vitro translation system is supplemented with individually purified GADD34 and K3L, the high levels of eIF2α phosphorylation seen in the HeLa extracts (Fig. 3B)  The overexpression of GADD34Δ and K3L prior to cell lysis helps alleviate the inhibitory effect of eIF2α phosphorylation and supports high levels of translation exemplified by the formation of robust polysomes (Fig. 1F). Furthermore, cellular expression of these two proteins increases the reproducibility of the in vitro translation experiments by decreasing the initial phosphorylation of eIF2α upon lysis (Fig. 3A). Although our translational system is not treated with micrococcal nuclease to deplete endogenous mRNAs, the extract preserves the integrity of all translation machinery components resulting in high levels of translation (Fig. 2, Fig. S1), and can be used with both m 7 G-cap dependent (HBB) and IRES dependent mRNAs (EMCV IRES) (Fig. 1G). The physiological state of the eIF2α may allow the study of the effect of different factors known to influence mRNA translation, including the roles of translation initiation factors and the effects of RNA sequence and secondary structure on translation initiation (49)(50)(51)(52)(53)(54)(55)(56)(57). The fact that the present CFPS system produces polysomes in a robust manner may also allow in vitro studies of the mRNA circularization model of translation (58,59). Finally, maintaining the physiological state of the eIF2α should also make the human-derived CFPS system valuable for assessing translational fidelity during alternative decoding events (37), programmed ribosomal frameshifting (60), and ribosome quality control pathways (28, 61).
The CFPS system described here provides substantial improvements to the use of in vitro cellular lysates for the study of human translation. However, future improvements are likely possible within this system. For example, it could be useful to find ways to deplete endogenous mRNAs from the extract, i.e. to eliminate the formation of residual disomes and trisomes (Fig.   4A) without the use of nucleases. Additionally, we observe phosphorylation of elongation factor eEF2 in all the extracts we examined (Fig. 4B). While additional biochemical analysis is required, the extracts might be further improved by decreasing eEF2 phosphorylation at residue T56 by using eEF2 kinase inhibitors (62). The approaches described here could be also used to prepare in vitro cell-free translation systems from other cell types, including immortalized cells derived from different tissues. It may also be possible to prepare CFPS systems from primary cells derived from healthy individuals as well as from patients suffering from diseases associated with translation defects, including rare ribosomopathies (63,64). Efforts to reduce the number of cells needed to prepare the engineered CFPS system described here would be required but could provide unprecedented opportunities to study the molecular mechanisms underlying human translation regulation.

Materials and Methods
Cloning humidity. Induction of GADD34Δ and K3L expression prior to cell extract preparation is described below.

Generation of the HEK293T cell line with a stably integrated GADD34 Δ1-240/K3L
construct. A stable cell line for endogenous expression of GADD34 Δ1-240 (GADD34Δ) and K3L was generated following the procedure described previously (24). Using Amaxa SF Cell Line 4D-Nucleofector X Kit (Lonza), WT HEK293T cells were transfected with a transposase-encoding plasmid (pSB100X), and the pSBtet-GADD34Δ/K3L construct (see above). After two days of recovery, nucleofected cells were selected with 300 µg/ml hygromycin (Thermo Scientific). A total of three rounds of selection were performed, each with approximately three doublings. Visual observation of the cell culture monitored the efficiency of the selection. After completion of the selection, cells were regularly maintained with 150 µg/ml hygromycin.
In vitro transcription reactions. In vitro transcription reactions were performed using PCR products generated with primers encoding a flanking T7 RNA polymerase promoter and a poly-A tail. Reactions were set up, as previously described (66)

Competing interests
The authors declare no competing interests.  Extracts were treated with the indicated concentrations of micrococcal nuclease prior to carrying out translation reactions with nanoluciferase mRNA. All error bars represent one standard deviation of three independent replicates.