Cell-free phosphoprotein synthesis of sfGFP

Our goal was to develop a CFPS platform capable of rapidly synthesizing an active human kinase. To accomplish this, we first optimized conditions that enabled site-specific incorporation of Sep into the superfolder GFP (sfGFP) reporter protein at a single in-frame amber codon at position 2 (sfGFP-S2TAG). Notably, the sfGFP gene sequence also encoded a C-terminal Strep-tag to enable purification and proteomics analysis. To assess phosphoprotein production in CFPS, 15 μl scale batch reactions were conducted for 20 h at 30 °C using the PANOx-SP system27,28. Combined transcription and translation in the CFPS reaction were driven by either the wild-type sfGFP or the sfGFP-S2TAG DNA template. Negative controls lacking DNA templates were also conducted. Using this approach, we synthesized 686±48 μg ml−1 of wild-type sfGFP and 567±37 μg ml−1 sfGFP-S2TAG in batch CFPS reactions that used extracts containing the overexpressed Sep-OTS (Fig. 1c, Table 1; mean±s.d.; n=4).

Table 1: Comparison of CFPS titres in this study relative to previous works. Full size table

Affinity-purified sfGFP was analysed by SDS–polyacrylamide gel electrophoresis (PAGE) to confirm production of full-length protein (Supplementary Fig. 1). We then digested the affinity-purified sfGFP with trypsin and confirmed incorporation of Sep at sfGFP-S2TAG by mass spectrometry (MS) using label-free shotgun proteomics. A representative tandem MS spectrum from a N-terminal tryptic peptide of sfGFP reporting incorporation of Sep at position S2 of sfGFP is provided in Fig. 1d. To assess the purity of modified sfGFP, we performed label-free quantitation using two strategies: spectral counting and MS1 intensity-based quantitation (Supplementary Tables 1–3, Supplementary Figs 2 and 3). Both approaches confirmed that the majority of sfGFP expressed with the Sep-OTS system contained Sep incorporated at position 2 of sfGFP-S2TAG. Near-cognate suppression via incorporation of amino acids Q, Y and K was responsible for read-through of the TAG codon as we had observed in our previous studies12,25 (Supplementary Table 1). We also found low levels of glycine read-through of the TAG codon. However, the incorporation of these natural amino acids was dramatically reduced in the presence of the Sep-OTS (Supplementary Figs 2 and 3). Last, the high sensitivity of our MS assay allowed us to detect direct evidence for ribosomal skipping at the TAG codon (Supplementary Tables 1–3). Relative intensities of peptides harbouring ribosomal skipping were 12.2 and 2.1% in the absence or presence of the Sep-OTS, respectively (Supplementary Table 3 and Supplementary Fig. 3).

Prior studies have suggested that the efficiency of amber codon suppression may be position dependent22,29,30. In other words, the position in which the nsAA is incorporated affects protein synthesis yields and incorporation efficiency. To study this effect, we further examined Sep incorporation in sfGFP at position E17. While sfGFP-E17TAG yields remained high at 516±9 μg ml−1 (Table 1), efficiency of Sep incorporation as determined by MS1 intensity-based quantitation was lower, ∼20% for sfGFP-E17TAG versus ∼90% for sfGFP-S2TAG (Supplementary Table 3). Similar to our results with S2, near-cognate suppression via incorporation of amino acids Q, Y, K and G was also observed at position 17 (Supplementary Table 1, Supplementary Fig. 4). Our results provide evidence for the positional dependence of nonsense suppression efficiency.

Since we intended to study a human kinase with two Sep residues (see below), we also studied site-specific phosphorylation into sfGFP at positions S2 and E17. As compared with wild-type sfGFP expression, yields were lower when synthesizing sfGFP-S2TAG/E17TAG (289±21 μg ml−1; Table 1). Unfortunately, efforts to detect the doubly phosphorylated sfGFP by MS failed due to technical limitations. Specifically, addition of multiple phosphorylated residues into peptides reduces their ionization efficiency. In addition, we speculate that the addition of two negative charges into the reporter peptide shifts the charge state of the peptide to a singly charged peptide with an m/z ratio outside of our detection window. To verify Sep incorporation in doubly phosphorylated sfGFP and validate both mono-phosphorylated forms, we used a Phos-tag-mediated mobility shift assay that has been previously described31,32 (Supplementary Fig. 5). The Phos-tag reagent binds phosphorylated proteins and causes slower migration of phosphorylated proteins during SDS–PAGE. Using the Phos-tag method, we observed distinct signature shifts for phosphoserine-containing fractions of S2TAG, E17TAG and S2TAG/E17TAG sfGFP preparations compared with the unmodified, wild-type sfGFP (Supplementary Fig. 5). Overall, Phos-tag and MS data validated a novel CFPS platform that permits biosynthesis of proteins harbouring site-specifically incorporated Sep residues.

Cell-free phosphoprotein synthesis of MEK1 kinase

We next demonstrated the utility of our method by applying the platform for the in vitro synthesis of highly active doubly phosphorylated human MEK1 kinase (mitogen-activated ERK activating kinase 1). In previous works, bacterial expression of soluble MEK1 in vivo has required fusion partners, such as the maltose binding protein9 or glutathione S-transferase33 fusion tags. Since CFPS systems have shown benefits for expressing mammalian proteins in soluble form as compared with in vivo expression platforms34, we attempted to express MEK1 without fusion partners. Wild-type MEK1 with serines at positions S218 and S222 (MEK1-SS) and doubly phosphorylated MEK1 with phosphoserines at the same positions (MEK1-SPSP) were produced in 20-h CFPS batch reactions at 30 °C (Fig. 2a). A unique feature of our technology for co-translational incorporation of Sep is the ability to produce mono-phosphorylated forms of kinases without having to add alanine mutations to suppress second site phosphorylation events. We therefore also synthesized MEK1 variants phosphorylated at either S218 (MEK1-SPS) or S222 (MEK1-SSP) having the native activation loop amino-acid sequence to determine if each site individually could activate the kinase alone. The total MEK1 expression yields for each variant were quantified by 14C-leucine incorporation. We observed synthesis of 308±19 μg ml−1 MEK1-SS, 343±9 μg ml−1 MEK1-SPS, 328±36 μg ml−1 MEK1-SSP and 269±28 μg ml−1 of MEK1-SPSP (Table 1, Fig. 2a). These volumetric yields (g l−1) exceeded previously reported in cell produced MEK1 by >1,000-fold (Table 1).

Figure 2: In vitro synthesis of phosphorylated MEK1 variants. (a) Volumetric yields of phosphorylated MEK1 variants MEK1-SS, MEK1-SPS, MEK1-SSP and MEK1-SPSP produced in 15 μl batch reactions. (b) Time-course plot showing the biosynthesis of MEK1-SPSP in 300 μl batch reactions. The inset shows a representative autoradiogram of MEK1-SPSP being synthesized over time. (c) Quantitation of total MEK1 production and phosphoprotein production by western blot analysis on SDS–PAGE and Phos-tag gels. The Phos-tag western blot shows characteristic shifted bands for the phosphorylated MEK1 variants MEK1-SPS, MEK1-SSP and MEK1-SPSP. These bands are absent in the wild-type MEK1-SS control sample. Error bars represent s.d. from three independent samples. N=3 for all western blot and gel analysis. Full size image

While phosphoprotein production in 15 μl batch reactions provides throughput, we next set out to demonstrate the potential for scale-up, noting that several previous works have demonstrated the ability to scale CFPS to the litre scale35,36. For example, Yin et al.36 used the open reaction environment of CFPS to produce 300 μg ml−1 aglycosylated trastuzumab in reactions ranging from 60 μl to 4 l. By increasing the batch reaction scale more than 10-fold to 300 μl reactions in a 24-well flat-bottom plate, we observed production of 467±15 μg ml−1 MEK1-SPSP (Fig. 2b). The increase in productivity from our 15 μl reactions is consistent with previous reports37,38, which have noted that increasing the surface area-to-volume ratio can increase CFPS yields. We pooled together eight CFPS reactions to make 1.05±0.12 mg of total MEK1-SPSP as determined by radioactive incorporation (Fig. 2b). The capacity to manufacture milligram quantities of site-specifically phosphorylated proteins demonstrates that our approach is not restricted to microgram quantities for analytical purposes.

Having successfully expressed different forms of MEK1, we next characterized the protein products in several ways. First, we validated the production of full-length MEK1 by western blot analysis using an antibody specific for the C-terminal His tag (Fig. 2c; full gels available in Supplementary Fig. 6). As expected, full-length MEK1, MEK1-SPS, MEK1-SSP and MEK1-SPSP were produced when the DNA template contained site-specific in-frame amber codons. Second, we confirmed MEK1-SPSP production by western blot using a phosphospecific antibody for MEK1 (Supplementary Fig. 7). Third, we validated the presence of Sep using the Phos-Tag gel-shift assay, because similar to doubly phosphorylated GFP, we were unable to detect the MEK1-SPSP peptide by MS for technical reasons. The gel shifts for MEK1-SPS, MEK1-SSP and MEK1-SPSP were distinct, and as noted by others39, the doubly phosphorylated MEK1 protein counter-intuitively produced a characteristic shift that was intermediate when compared with the mono-phosphorylated MEK1 variants (Fig. 2c), an observation that is also consistent with the phosphorylated sfGFP (Supplementary Fig. 5). Our Phos-tag data provided clear demonstration of single- and two-site phosphorylation into MEK1. While our data suggest that the product is not completely pure (that is, there are some non-phosphorylated species), we believe that continued developments to build a more efficient Sep-OTS40,41 or removal or phosphatases from the extract will provide notable improvements in purity as seen for other OTSs. We then used our platform to elucidate sequence-function relationships for MEK1.

Functional effects of site-specific phosphorylation in MEK1

MEK1 plays an important role in cellular signal transduction through the human MAP kinase cascade responsible for driving cellular proliferation and differentiation42. Within the cascade, the doubly phosphorylated MEK1 phosphorylates the extracellular-signal-regulated kinases ERK1 and ERK2 (Fig. 3a). We therefore examined the enzymatic activity of cell-free synthesized mono- and doubly phosphorylated MEK1 variants towards ERK2 with an in vitro kinase cascade assay. To evaluate kinase activities, we incubated the CFPS reaction products MEK1-SS, MEK1-SPS, MEK1-SSP or MEK1-SPSP with a full-length ERK2 K54R variant with greatly reduced autophosphorylation activity. Western blot analysis with an anti-phos-ERK antibody showed that the doubly phosphorylated MEK1 could robustly phosphorylate ERK2 in vitro as expected (Fig. 3b; full gels available in Supplementary Fig. 7). Uniquely, our platform also enabled us to test the functional role of mono-phosphorylated forms on MEK1 activity. We discovered that phosphorylation at either S218 or S222 is necessary and sufficient for MEK1 kinase activity (Fig. 3b). To our knowledge, this is the first demonstration that three different forms of MEK1, with native phosphoserine, produce active kinase. It is also the first demonstration of active MEK1 synthesis by bacterial CFPS without the need for fusion proteins to enable soluble protein expression. Obviating fusion proteins during the production of human proteins in CFPS systems as demonstrated here will provide new opportunities to study the kinase without confounding effects of the solubility protein partner. The flexibility and utility of the CFPS system is shown by the rapid, robust and scalable production of the active phosphorylated MEK1-SPSP for biochemical and biophysical characterization.