Sandwiched‐fusion strategy facilitates recombinant production of small labile proteins

Abstract

Efficient production of large quantities of soluble, properly folded proteins is of high demand in modern structural and functional genomics. Despite much advancement toward improving recombinant protein expression, many eukaryotic proteins especially small peptides often fail to be recovered due to rapid proteolytic degradation. Here we show that the sandwiched‐fusion strategy, which is based on two protein tags incorporated both at the amino‐ and carboxyl‐terminus of target protein, could be employed to overcome this obstacle. We have exploited this strategy on heterologous expression in Escherichia coli of eight small degradation‐prone eukaryotic proteins, whose successful recombinant productions have yet to be achieved. These include seven mitochondria‐derived peptides (MDPS), a class of unique metabolic regulators of human body, and a labile mosquito transcription factor, Guy1. We show here that the sandwiched‐fusion strategy, which provides robust protection against proteolysis, affords an economical method to obtain large quantities of pure five MDPs and the transcription factor Guy1, in sharp contrast to otherwise unsuccessful recovery using the traditional amino‐fusion method. Further biophysical characterization and interaction studies by NMR spectroscopy confirmed that the proteins produced by this novel approach are properly folded into their biologically active structures. We anticipate this strategy could be widely utilized in production of other labile protein systems.

1. INTRODUCTION

Cost‐effective production of functional proteins remains a major obstacle in studies of structural and functional proteomics. Currently Escherichia coli remains the most popular host for recombinant protein expression, largely owing to its advantageous, inexpensive and high yield protein production, the well‐characterized genetics, ease of manipulation, and a wide variety of available molecular tools. However, expression of heterologous proteins in the bacterial system frequently encounters many challenges which must be surmounted. These challenges include: (a) the lack of sophisticated machinery to perform posttranslational modifications and molecular chaperones that help protein folding, which results in misfolded protein with aberrant biological activity or inclusion body formation; (b) toxicity issues that lead to low growth rate and even cell death; (c) inefficient protein translation caused by conflict in genetic codon usage; and (d) rapid proteolytic degradation of the target protein that is viewed as unwanted by cells. Although no universal approach has been established for efficient recombinant expression of any given protein, much advancement has been achieved to circumvent the above mentioned obstacles, including introduction of genetic modified E. coli strains, ¹ co‐expression with molecular chaperone systems, ² usage of strong promotors, and gene fusion systems, ³ and so on. Among the approaches mentioned above, gene fusion technology has been shown to be the most effective strategy to improve the folding and solubility of recombinant proteins, to simplify procedures of protein isolation, and to protect protein of interest from proteolytic degradation. ⁴ A variety of fusion tags has been developed, such as the widely used affinity tag His‐tag that facilitates the following purification steps, and a large number of solubility enhancer tags including glutathione S‐transferase (GST), maltose‐binding protein (MBP), thioredoxin, the B1 domain of Streptococcal protein G (GB1), ubiquitin and SUMO, and so on; ⁵ and new solubility enhancer tags are constantly emerging.

Despite the wealth of well‐established approaches, in many cases recombinant production of eukaryotic proteins remains challenging, especially for the ones of small size that are often susceptible to severe proteolysis. Mitochondrial‐derived peptides (MDPs), which are encoded by the small open reading frames (sORFs) within other known genes of the mitochondrial DNA, serve as novel circulating signaling molecules for organism cytoprotection and energy regulation. ⁶ , ⁷ This class of small peptides, ranging from 20 to 38 amino acids, comprise humanin and six small humanin‐like peptides. As the first MDP identified, humanin has been found to play a diverse role in many biological processes including apoptosis, cell survival, substrate metabolism, inflammatory response, and response to stressors such as oxidative stress, ischemia, and starvation. ⁸ , ⁹ Similar to humanin, the small humanin‐like peptides also confer cytoprotective effects and enhance mitochondrial metabolism via increasing oxygen consumption rate, although each SHLP may differentially regulate mitochondrial and cellular health and functions. ¹⁰ , ¹¹ Due to the beneficial effects, MDPs have been subjected to intensive study to elucidate the mechanism that how these peptides serve as signals for organism cytoprotection and energy regulation, as well as therapeutic exploitation in many human diseases including Alzheimer's disease, stroke, obesity, diabetes, myocardial ischemia and reperfusion, atherosclerosis, amyotrophic lateral sclerosis, and certain types of cancer. ⁶ , ⁷ , ⁹ , ¹¹ , ¹² , ¹³ Since the discovery of MDPs nearly two decades ago, to date, however, all MDPs used for research were chemically synthesized; and the structural and biochemical studies have been greatly hampered by the high cost. More importantly, due to the small size and inherent dynamic nature, NMR spectroscopy offers an extremely powerful tool for investigating the structures and protein–protein interactions of MDPs. However, the daunting cost that is associated with chemical synthesis of NMR sensitive isotope ( ¹⁵ N or ¹³ C) enriched peptide has made it unfeasible for NMR applications. This has emphasized the need for an efficient and low‐cost method for producing MDPs.

Here we present a robust strategy for recombinant production of MDPs in E. coli. In fact, we initially attempted to overexpress MDPs in E. coli with protein tags incorporated at the amino‐terminus. However, despite various protein tags and expression systems were tested, isolation of MDPs was unsuccessful due to severe proteolytic degradation occurred during purification process. To circumvent this situation, we have exploited a tripartite fusion strategy, hereafter referred as the “sandwiched‐fusion” system, in which the gene of interest is fused between two different protein tags. In fact, the tripartite fusion concept was originally developed more than three decades ago to facilitate purification of full‐length human insulin‐like growth factor II out of the pool of degraded fragments. ¹⁴ In that work, two affinity fusion tags, an IgG‐binding ZZ domain and an HSA‐binding region of protein G (B1B2), were incorporated at the amino‐ and carboxyl‐terminus, respectively, to allow a two‐step affinity chromatography workflow, which ensured the purity of target protein. Here, we replaced the two affinity fusion tags with MBP and GB1, two fusion tags that have been widely used for recombinant protein production. ⁵ Coupled with overexpression in E. coli, this fusion strategy, which provides robust protection from proteolysis, affords an economical method to obtain large quantities of pure five MDPs in a fast and efficient manner. Multiangle light scattering (MALS) analysis reveals all five MDPs adopt a monomeric conformation in aqueous solution, whereas circular dichroism (CD) analysis demonstrates that various MDPs contain distinct secondary elements. Moreover, production of isotope enriched MDPs samples is straightforward as the labeling schemes have been well‐established in E. coli. Thus, we further prepared the ¹⁵N‐labeled samples of Humanin and other MDPs in aqueous solution for the first time, and collected the well‐resolved ¹⁵ N‐HSQC 2D spectra. Subsequent NMR titration experiments confirmed the interaction of humanin peptide with both major isoforms of the molecular chaperone Hsp90, indicative of proper fold of MDP samples prepared by this novel approach. Notably, we show here the in vitro evidence that the interaction pattern is preserved for another molecular chaperone, Hsc70. The versatility of this novel strategy was further demonstrated by successful production of Guy1, a sex‐determining transcription factor from mosquito,¹⁵ indicative of the potential of application of this sandwiched‐fusion strategy to other labile protein systems.

2. RESULTS

2.1. The novel sandwiched‐fusion system allows recombinant production of humanin in E. coli

Heterologous expression primarily in bacteria for the production of recombinant proteins remains the workhorse for scientists in academics and in pharmaceutical companies for structure‐based drug design studies. Despite the biological significance and potential pharmaceutical applications of MDPs, to our knowledge, to date there has been no report of recombinant production of any of these novel signaling peptides, which has greatly hampered a better understanding of their mechanisms of action and subsequent structure–function analyses. Because MDPs often contain a high proportion of hydrophobic residues, to improve solubility as well as protein yield, initially we tried to express the humanin gene using a conventional amino‐terminal fusion strategy. Briefly, the humanin genes were synthesized on the basis of codon preference in E. coli and inserted into a pET16‐derived vector with the poly‐His affinity tag followed by a MBP tag at the amino‐terminus (Figure 1d). Typical purification results are demonstrated in Figure 1a, which shows that although after IPTG induction high level of expression is achieved (lane 2 in Figure 1a), recovery of humanins is greatly hampered by complications related to rapid proteolysis occurred during purification step (lane 5). Indeed, due to the proteolysis occurred during purification process, in the end we failed to recover any protein of interest except for the amino‐terminal fusion tag, even though different host E. coli strains, various protease inhibitors and purification procedures were tested.

FIGURE 1.

Preparation of the humanin peptide using the sandwiched‐fusion strategy from E. coli. (a) A representative SDS‐PAGE gel showing comparison of expression and purification of humanin using the traditional amino‐terminal bipartite fusion strategy, with that using the sandwiched‐fusion tripartite strategy. Lanes 1 (3), 2 (4), 5 (7), 6 (8) denote the samples of His₆‐MBP‐TEV‐humanin for before IPTG induction, after induction, Ni‐NTA purification and TEV protease treatment, respectively, whereas the lane numbers in parentheses denote the corresponding samples prepared using the tripartite‐fusion construct of His₆‐MBP‐TEV‐3C‐GB1. It is noted that lane 5 shows two major bands, whose molar masses match that of the MBP‐humanin fusion and MBP tag only, respectively. This observation suggests the sample had been subjected to severe proteolytic degradation during the purification process, which eventually led to failure of recovery of target protein after TEV cleavage (lane 6). In contrast, lane 7 shows only one major band, suggesting the target protein has been efficiently protected from degradation by the sandwiched‐fusion strategy. Lane 9 shows the humanin‐GB1 fusion purified after SEC. PM: protein marker. (b) The purified humanin‐GB1 adopting a stable conformation in solution as evidenced by the symmetric SEC profile. (c) Amino acid sequence of humanin. (d) The construct scheme of the conventional amino‐terminal bipartite fusion system. (e) The construct scheme of the sandwiched‐fusion system. (f) The construct scheme of the carboxyl‐terminal fusion system. (g) SDS‐PAGE gel showing the preparation of the humanin peptide using the carboxyl‐terminal fusion system. The samples were loaded in the same order as the SDS‐PAGE gel shown in (a)

In order to overcome the proteolysis problem, we designed the sandwiched‐fusion construct with the idea that the carboxyl‐terminal fused protein tag may sterically hinder proteolysis from cellular protease and thus provides effective protection. As shown in Figure 1e, the gene encoding the target protein was inserted between two flanking genes encoding the MBP and GB1 protein tags, respectively. The amino‐terminal polyHis‐tag facilitates the purification process on a large scale using immobilized metal‐affinity chromatography (IMAC), while two protease cleavage sites were engineered between fusion‐tags and gene of interest for convenience of subsequent tag removal. Moreover, a six‐residue (GSGSGS) GS‐linker is engineered between the target gene and the carboxyl‐end GB1 fusion tag to further introduce flexibility and thus minimize potential interference from the fusion tag.

To prepare the humanin sample, E. coli cells containing the engineered sandwiched‐fusion plasmid was/were grown overnight after IPTG induction, and the soluble fraction was collected by osmotic shock followed by SDS‐page analysis. The high expression level of humanin is evidenced by a major band corresponding to the full‐length fusion protein of His₆‐MBP‐Humanin‐GB1 (lane 4 in Figure 1a). In the next step we used nickel (Ni²⁺)‐affinity chromatography together with size exclusion chromatography (SEC) to purify the humanin peptide. A typical procedure for purification is described in detail in the methods section. Briefly, as the majority of the fusion protein partitions into the soluble fraction, the supernatant clarified by centrifugation was next passed through a Ni²⁺‐chelating column. It is noteworthy that comparing to the bipartite humanin fusion sample that is highly susceptible to degradation (see the multiple bands shown in lane 5 of Figure 1a, which suggests the target protein is subjected to proteolysis), the sandwiched tripartite fusion strategy provides sufficient protection from proteolysis (see only one major band in lane 7 of Figure 1a). Protein eluted from the column was then subjected to TEV protease cleavage for removal of the amino‐terminal MBP tag as well as the poly‐His affinity tag. In principle, the human rhinovirus (HRV) 3C protease can also be added to detach the carboxyl‐terminal GB1 tag. However, we found that humanin peptide aggregates upon the cleavage of GB1 tag, which is not surprising as it contains a large portion of hydrophobic residues (Figure 1c). The highly water soluble GB1 in this scenario behaves as a solubility tag ensuring MDPs solubilized under aqueous conditions. Due to its small size, as well as overwhelming evidences showing that the presence of GB1 often does not interfere the structure of its fusion partners, ¹⁶ we decided to keep this solubility tag for following characterizations. After TEV protease cleavage the protein mixture was re‐applied to a Ni–NTA column to subtract polyHis‐tagged MBP and TEV protease, and the unbound humanin‐GB1 fusion protein was further subjected to a SEC column. The elution profile of SEC demonstrates a single major peak for humanin‐GB1 with a symmetric shape, indicative of a homogeneous conformation in aqueous solution (Figure 1b). A typical full SEC profile of humanin‐GB1 fusion prepared using the sandwiched‐fusion strategy is shown in Figure S1a. The purity of purified humanin‐GB1 fusion protein sample was assessed by SDS‐PAGE analysis (lane 9 in Figure 1a).

Having established vital role of the fusion tag at the carboxyl‐terminus in protecting of the target protein from proteolytic degradation, in a next step, we asked whether the amino‐terminal MBP tag is necessary in production of humanin. We therefore prepared a construct of His₆‐humanin‐GB1 with a TEV protease site inserted between the genes of His₆‐tag and humanin (Figure 1f), and compared the purification results. We found that although humanin‐GB1 fusion protein could be recovered and eventually purified using the carboxyl‐terminal fusion strategy (Figure 1g, lane 5), most recombinant protein formed soluble aggregates and eluted at the void volume of size‐exclusion chromatography (Figure S1b). As a result, the yield of properly folded humanin‐GB1 was only 2.2 mg per liter of LB media, ~28% of that achieved by the sandwiched‐fusion method, in line with the well‐established role of MBP as a solubility tag that enhances protein stability.

In summary, we show here that, in sharp contrast to no protein of interest recovered from the amino‐terminal fusion construct, the sandwiched tripartite fusion system both affords efficient protection to humanin from proteolysis and ensures high yields of protein with proper fold.

2.2. Application of the sandwiched‐fusion strategy for recombinant production of other MDPs

We next tested the sandwiched‐fusion strategy on recombinant production of six other MDPs, that is, the small humanin‐like peptides SHLP1‐6 (Figure 2a). As shown in Figure 2b, large amount of all six MDPs were recombinantly produced in E. coli by following the same protocol, and four MDPs, that is, SHLP1, SHLP2, SHLP5 and SHLP6, were successfully isolated with high purity as evidenced by the SDS‐PAGE analysis (Figure 2b). However, we found a large portion of SHLP3‐GB1 and SHLP4‐GB1 peptides formed soluble aggregation after removal of the amino terminal His₆‐MBP tag under the buffer condition used here, as evidenced by the elution volume at the void volume of SEC profile. Further work of buffer optimization is required to sustain native conformations for these two MDPs. As a result, both SHLP3 and SHLP4 are excluded here from further characterizations.

FIGURE 2.

Overexpression and purification results of six MDPs using the sandwiched‐fusion strategy. (a) Amino acid sequences of six MDPs. (b) Representative SDS‐PAGE gels showing purification results of the six MDPs as labeled. For each MDP, lanes 1–5 denote the samples of before IPTG induction, after induction, Ni‐NTA purification, after TEV protease cleavage, and SEC purification, respectively. SHLP3 and SHLP4 formed aggregate after TEV protease treatment. PM: protein marker. (c) Protein yields of five MDPs using the sandwiched‐fusion strategy

Taken together, the tripartite sandwiched‐fusion strategy allows recombinant production of all seven MDPs in E. coli, and the purification procedure presented here enables us to prepare five MDP samples with both high purity and homogenous conformations. The yields of each SEC‐purified MDP‐GB1 protein per liter of bacterial Luria Broth (LB) culture and M9 minimal media are listed in Figure 2c, based on more than three consecutive protein expressions and purifications using the same protocol. These quantities are adequate for conducting detailed functional and structural studies, for example, conducting crystallization screening or using heteronuclear NMR measurements.

2.3. Biophysical characterization of MDPs

To date, out of seven MDPs, only the chemically synthesized humanin and its variants have been biophysically characterized. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² To explore the biophysical properties of MDPs recombinantly prepared using the sandwiched fusion strategy, we first performed size‐exclusion chromatography (SEC) coupled to both MALS and quasi‐elastic light scattering (QELS) to assess the oligomerization status of MDPs. As shown in Figure 3, all five MDPs adopt a monomeric conformation in aqueous solution. The monomeric state of Humanin‐GB1 is consistent with the results reported previously for the chemically synthesized peptide. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²²

FIGURE 3.

SEC‐MALS analysis of various MDP‐GB1 fusions. (a) SEC‐MALS profiles showing all five MDP‐GB1 fusions adopt a single oligomeric state in solution. (b) The molar masses measured by MALS analysis are in reasonable agreement with the theoretical values, indicating all five MDPs being monomeric in solution

We next utilized the far‐UV circular dichroism (CD) spectroscopy to probe the secondary structure contents of MDPs. As all MDPs were acquired as MDP‐GB1 fusions, the CD signals of MDPs were further processed by subtraction of that of GB1 tag alone as baseline. The resulted CD spectra of five MDPs after GB1 signal subtraction are graphically depicted in Figure 4a. The CD spectra of humanin, SHLP2 and SHLP6 are characterized by two deep ellipticity minima at ~209 and ~220 nm, respectively, indicative of that these three MDPs adopt predominantly helical conformations in solution. The peptide SHLP1 also demonstrate a similar spectra shape with two ellipticity minima CD profile. However, the magnitude of negative molar ellipticity is drastically smaller, diagnostic of a lower percentage of helical structural elements existing in SHLP1. In contrast, the CD spectrum of SHLP5 presents a pronounced ellipticity minimum at ~199 nm, indicative of high percentage of disordered conformation in solution. ²³ Estimations of the secondary structure elements of all five MDPs were performed using the CDNN program equipped with CD spectrometer and are shown in Figure 4b. Notably, we show here that the humanin primarily adopts a helical structure under the condition measured here, different from the observation made using the synthesized peptide, which showed humanin being primarily unstructured in aqueous solution. ¹⁶

FIGURE 4.

Secondary structure characterization of MDPs by far‐UV CD spectroscopy. (a) The resulted CD spectra of MDPs obtained by subtraction of the CD signals of GB1 fusion tag from that of the MDP‐GB1 fusion proteins. CD spectra of GB1 and MDP‐GB1 fusions were collected at the concentration of 15 μM at room temperature. (b) Deconvolution of CD spectra of five MDPs showing their secondary structure elements with percentage

NMR is an extremely robust technique to monitor the conformation of a protein sample, and the 2D {¹H, ¹⁵N} HSQC spectrum is often referred as the fingerprint region of a protein. We thereby prepared the samples of ¹⁵N‐enriched MDP that fused to GB1 and collected {¹H, ¹⁵N}‐ heteronuclear single‐quantum coherence (HSQC) NMR spectra. As demonstrated in Figure 5, all five MDP‐GB1 samples yield high‐quality NMR spectra with good spectral dispersion and largely uniform peak intensities. Importantly, the number of well‐resolved peaks is approximately equal to that expected for each corresponding MDP‐GB1 fusion construct. The observation of a single set of NMR resonances readily suggests that these five MDP‐GB1 samples adopt a stable conformation in solution, in line with the results of MALS analysis shown in Figure 3. It is worth mentioning that further NMR spectra overlay revealed that comparing to the HSQC spectrum of GB1 along, only marginal resonance shifts were observed for the GB1 tag fused to MDPs, whereas most peaks remain unaffected (Figure S1). As this region is extremely sensitive and any interaction would result in perturbation in the chemical shifts from the original positions or changes in resonance intensities, we conclude that the MDP and its fused partner GB1 behave as two individual entities in solution and the potential interference introduced by protein fusion with the MDPs properties is kept to minimum, consistent with a plethora of previous results.¹⁶ The dispersion of NMR signals for MDPs presented in Figure 5 is in good agreement with the results obtained by CD spectroscopy. The high‐quality spectra enable future structural and in vitro functional studies using the NMR technique.

FIGURE 5.

{¹H, ¹⁵N}‐HSQC NMR spectra acquired for isotopically labeled MDP‐GB1 fusions prepared by the sandwiched‐fusion strategy. (a) Humanin‐GB1; (b) SHLP1‐GB1; (c) SHLP2‐GB1; (d) SHLP5‐GB1; and (e) SHLP6‐GB1. All spectra were acquired with sample concentrations of 100 μM at 283k. The high quality of spectra, which are evidenced by largely uniform signal intensities and corresponding numbers of peaks, lays solid foundation for future structure–function studies by NMR spectroscopy

2.4. MDPs recombinantly produced by the sandwiched‐fusion strategy are properly folded

To test whether the MDP samples prepared using the sandwiched‐fusion strategy are folded into biologically active conformations, in a next step we chose to carry out interaction studies of Humanin, the best‐studied MDP. It has been demonstrated that Humanin functions as an endogenous activator of chaperone‐mediated autophagy (CMA) pathway and exerts its cytoprotective effects via interaction with a key molecular chaperone, Hsp90. ²⁴ We thus prepared the ¹⁵N‐labeled Humanin and performed NMR titration with two major cytoplasmic isoforms of Hsp90, the inducible form Hsp90α and the constitutive form Hsp90β. As shown in Figure 6, addition of either Hsp90 isoform causes dramatically broadening effects specifically to a few residues. These observations thus confirm the direct interaction of humanin peptide with the chaperone Hsp90, and readily suggest the humanin sample prepared by the sandwiched‐fusion system retains its biological activity. Interestingly, Hsp90α and Hsp90β exerted literally identical binding effects to humanin (Figure 6a,b), indicating there is no difference for the two highly‐conserved Hsp90 isoforms in interacting with this signal peptide. In order to further identify the recognition sites of Hsp90s in humanin, we prepared ¹⁵N, ¹³C‐enriched humanin sample and recorded a series of 3D NMR data for backbone NMR resonance assignment. Completeness of assignment allows us to pinpoint the residues that beard dramatic NMR line broadening effect upon addition of Hsp90. Differential line broadening analysis ²⁵ shows there is only one Hsp90‐binding region in humanin spanning from residue 11 to 18 (Figure 6d). Intriguingly, there seems no correlation between the hydrophobicity of the humanin sequence and the binding preference by Hsp90 (Figure 6e), different from the client recognition mechanism of some other chaperone systems. ²⁶ , ²⁷ The detailed structural and mechanistic studies of Hsp90‐humanin are currently underway in our lab.

FIGURE 6.

Interactions assessed by {¹H, ¹⁵N}‐HSQC NMR titrations of 100 μM ¹⁵N isotopically labeled humanin‐GB1 peptide with 100 μM molecular chaperones Hsp90α (a), Hsp90β (b) and Hsc70 (c). The signal broadening effect for peaks upon addition of chaperones indicates binding to humanin, which confirms the humanin sample prepared by the sandwiched‐fusion system is biologically active. Notably, the same residues were affected upon addition of Hsp90α, Hsp90β and Hsc70, indicating the interaction pattern is preserved for three chaperones. (d) HSQC peak intensity ratios of Hsp90α‐bound humanin to free humanin. Considerable reductions in intensity indicate that Hsp90α binds to the region between amino acids 11 and 18. (e) Binding sites in humanin (region in shade) mapped by differential line broadening analysis and plot of the hydrophobicity of humanin as a function of its primary sequence. A hydrophobicity score (Kyte‐Doolittle, window = 3) higher than zero denotes increased hydrophobicity. No strict correlation between the binding preference and hydrophobicity score was observed, suggesting the interaction is not solely driven by hydrophobic contacts

Molecular chaperone Hsc70 is another core player implicated in the CMA machinery that targets client proteins for lysosomal degradation. ²⁸ To investigate whether Hsc70 can also interact with humanin, we monitored the titration effects by NMR the 2D {¹H, ¹⁵N} HSQC spectrum of humanin in presence and absence of Hsc70 (Figure 6c). Interestingly, spectral analysis reveals Hsc70 binds to the same 8‐amino‐acid region in humanin, that is, residues 11–18, reminiscent of the binding scenario of Hsp90s. This result is in line with the observation that for a certain client, the same interaction pattern is preserved for various chaperones. ²⁹ Moreover, the absence of a KFERQ‐like motif in the Hcp70‐binding region, a canonical Hsc70‐client recognition signature in CMA pathways, ²⁸ implies Hsc70 utilizes different mechanism in binding to humanin from clients targeted for degradation. The physiological significance of the interaction between humanin and chaperone Hsc70 remains elusive.

2.5. The sandwiched tripartite fusion system efficiently stabilizes other labile protein system against proteolytic degradation

To explore whether the sandwiched‐fusion method is a versatile strategy that can be applied to other labile protein systems, we tested this method on a male‐determining transcription factor of mosquito, Guy1, whose recombinant production by E. coli has yet to be reported. For comparison, we also made two traditional amino‐terminal fusion constructs of His₆‐MBP‐Guy1 and His₆‐GB1‐Guy1. Similar to our previous observations in MDPs purifications shown in Figure 1a, using the conventional bipartite fusion construct, although high level of expression was achieved after induction with the expected molecular weight (Figure 7a, lanes 2 and 4), subsequent purification only yielded the amino‐end protein tags due to rapid proteolysis for both constructs (see the major bands presented in lane 7 and lane 9 in Figure 7a with molecular weights corresponding to the His₆‐MBP and His₆‐GB1 tags, respectively). Accordingly, attempts for isolation of target protein after TEV protease cleavage had been unsuccessful (lanes 8 and 10 in Figure 7a). In contrast, the sandwiched‐fusion construct, which provides robust steric protection against proteolytic degradation (compare the multiple bands in lane 5 and primarily one major band in lane 7 of Figure 7a), enabled successful recovery of large quantities of pure protein of interest (lane 9 in Figure 7), reminiscent of the case of MDPs productions (Figures 1a and 2b). This observation, again, unequivocally demonstrates the necessity of applying the sandwiched‐fusion strategy in protecting of labile protein from degradation.

FIGURE 7.

The sandwiched‐fusion strategy enables recombinant production of Guy1 and following biophysical characterizations. (a) A representative SDS‐PAGE gel showing the necessity of application of the sandwiched‐fusion tripartite strategy in preparing recombinantly expressed Guy1 against proteolytic degradation. Lanes 1 (3), 2 (4), 7 (9) and 8 (10) show overexpression and purification of the bipartite fusion construct of His₆‐MBP‐TEV‐Guy1 (His₆‐GB1‐TEV‐Guy1) for the samples of that before IPTG induction, after induction, Ni‐NTA purification and TEV protease treatment, respectively. Lanes 5, 6, 11 and 12 were loaded in the same manner for the samples of the sandwiched‐fusion construct of His₆‐MBP‐TEV‐Guy1‐GB1. Lane 13 shows the Guy1‐GB1 fusion purified after SEC. Reminiscent of the purification results shown in Figure 1a, the sandwiched‐fusion strategy has shown to be essential for recovery of Guy1, which otherwise is prone to degradation during purification process. (b) SEC‐MALS profile showing Guy1 behaves as a monomer in solution. (C) {¹H, ¹⁵N}‐HSQC NMR spectrum of ¹⁵N‐isotopically labeled Guy1‐GB1 acquired with a concentration of 100 μM at 283k

As shown in Figure 7b, the molecular mass of Guy1‐GB1 fusion protein measured by SEC‐MALS was 14.3 kDa, in reasonable agreement with its theoretical value of 13.6 kDa, indicative of Guy1 being a monomer in solution. The protein yields for Guy1‐GB1 in LB and M9 minimal media were determined to be ~10 and ~8 mg per liter of culture, respectively. We further prepared the ¹⁵N‐enriched Guy1‐GB1 sample and collected the {¹H, ¹⁵N}‐heteronuclear HSQC data. As shown in Figure 7c, the NMR spectrum presents favorable dispersion and narrow line widths, which indicated that the Guy1‐GB1 adopts a single stable conformation under the experimental condition, laying the foundation for further NMR based structural characterization and functional studies. Taken together, the versatility of this novel sandwiched‐fusion strategy was demonstrated by successful purification of a labile transcription factor other than MDPs.

3. DISCUSSION

To date, the expression of heterologous genes in bacteria for production of recombinant proteins remains the workhorse for basic research or commercial purposes, primarily due to the simplicity of manipulation and low cost. In spite of numerous advancements that have been achieved, recombinant expression in bacteria frequently encounters many challenges, including codon bias, low level or no expression, protein insolubility, stability issue, protein inactivity, and host cell toxicity, and so on. Recent work has identified a novel component of the proteome: the translation of sORFs, which has demanded an examination of the structure and function of biologically active peptides that are encoded by sORFs and have regulatory roles in eukaryotic cells. ³⁰ For such proteins of small size which often lack a stable tertiary structure, proteolytic degradation remains a major obstacle, as recombinant peptides are often viewed as unwanted by cells and are subjected to rapid proteolysis, resulting in low or no yield of desired protein.

MDPs are a class of unique metabolic regulators of human body encoded by mitochondrial sORFs, playing an important cytoprotective role in maintaining mitochondrial function and cell viability under pressure. Therefore MDPs are considered to be novel biomarkers or therapeutic targets for treating many human diseases including cardiovascular disease, ⁶ Alzheimer's disease, prostate cancer, macular degeneration, and diabetes. ¹³ However, the absence of a methodology for efficient producing MDPs in a large quantity, and at the same time at low cost, has greatly impeded their biochemical characterization or their developments in pharmaceutical processes.

The current study sets out to establish a methodology for recombinant production of biologically active small degradation‐prone peptides including MDPs. Although at this moment it remains to be defined which E. coli protease is responsible for the degradation, we show that the sandwiched strategy, where the gene encoding the desired product is fused between two genes encoding two different fusion partners, provides necessary protection of the target protein from proteolysis and thus leads to remarkable increase in protein yield. The most prominent feature of the system is that the two flanking fusion partners, instead of one fusion tag in the conventional bipartite amino‐fusion system, provide steric protection to both terminals of the target protein to prevent proteolysis, and therefore ensure recovery of the desired protein during protein purification process. Moreover, the amino‐terminal tag offers stabilizing effect to the protein of interest, at least in the case of aggregation‐prone peptide such as the humanin tested here, and offers high yields of proteins that are properly folded. To the best of our knowledge, this work presented here provides the first detailed description on the production in E. coli of milligram quantities of five high‐purity functional MDPs, as well as a mosquito transcription factor, Guy1. Therefore, the present investigation suggests that this versatile strategy can be applied for recombinant production of many labile protein systems whose recombinant production cannot be achieved using a conventional bipartite fusion construct. Although we have chosen a sandwiched‐fusion system based on the MBP and GB1 here, we argue that other commonly‐used fusion tags such as SUMO, GST, and TRX may also be exploited. Moreover, after removal of the amino‐terminal MBP tag, all five MDP‐GB1 fusions, as well as the Guy1‐GB1 fusion, demonstrated high stability, as evidenced by nearly no change in the NMR spectral quality after at least 1 week, which is sufficient for most biophysical and biochemical characterizations. However, further cleavage of the carboxyl‐terminal GB1 fusion partner resulted in aggregation of humanin and Guy1, suggesting, at least in some cases, the presence of a solubility tag is necessary for stabilizing the peptide of interest in solution.

In fact, due to the lack of a robust strategy for recombinant production of MDPs, as of now all research related to structural characterizations of MDPs has limited to chemically synthesized humanin peptides, which were performed either under extremely acidic condition or in organic solvent of 30% trifluoroethanol (TFE).^17‐21 However, it has been well documented that TFE may artificially induce helix‐forming properties of peptides, ³¹ necessitating a detailed re‐examination of humanin properties, as well as investigations of other MDPs, in a context close to the physiological condition. Sufficient amounts of isotope‐enriched MDPs samples, which are supplied by the sandwiched‐fusion strategy presented here, enabled us to record a series of NMR spectra of excellent quality, laying solid foundations for further structural determination and structure–function studies by NMR spectroscopy.

4. MATERIALS AND METHODS

4.1. Chemicals

The target DNA genes applied to amplify DNA were synthesized by General Biol company. Primer synthesis and plasmid DNA sequencing were performed by TSINGKE Biological Technology. All nucleic acid purification kits were from Axygen Scientific Inc. Restriction enzymes were purchased from New England Biolabs.

Tryptone and yeast were from Thermo Scientific. Imidazole, Na₂HPO₄∙7H₂O, NaCl, Tris, KH₂PO₄, NH₄Cl, NaCl, MgSO₄, PMSF, CaCl₂, vitamin B1 (thiamine), Ampicillin, d‐(+)‐glucose were purchased from Sangon Biotech. Ethanol absolute, isopropanol, hydrochloric acid were from HUSHI. 2‐Mercaptoethanol (βme) was from Macklin. IPTG (isopropyl‐β‐d‐1‐thiogalactopyranoside) was from Biomiky, ¹³C glucose and ¹⁵NH₄Cl, applied to product isotopically labeled samples were obtained from Cambridge Isotope Laboratories Inc.

4.2. Fusion constructs

All recombinant bipartite fusion plasmids, including His₆‐MBP‐TEV‐MDPs, His₆‐MBP‐TEV‐Guy1, His₆‐GB1‐TEV‐Guy1, His₆‐GB1‐HSC70 and His₆‐TEV‐GB1, were constructed by modification of pET‐16a vector using the method of restriction enzyme digestion and ligation. Briefly, the PCR product of amplified insert gene and pET16a‐His₆‐MBP or pET16a‐His₆‐GB1 vector were double‐digested by restriction enzymes of NdeI and BamHI (NEB). Insert was then ligated to the corresponding vector using a DNA Ligation Kit (Takara) following gel extraction. Constructs of His₆‐TEV‐HSP90A and His₆‐TEV‐HSP90B were prepared by insertion of the amplified gene encoding the corresponding target protein into the pET‐16a vector without a fusion tag.

All tripartite fusion constructs were made by introducing a 3C‐GS‐linker (three GS repeats)‐GB1 at the carboxyl‐terminus of the corresponding bipartite fusion plasmid using ClonExpress II One Step Cloning Kit (Vazyme) according to the manufacturer's instructions. Recombination primers for amplified target genes were designed by the online CE Design software (https://crm.vazyme.com/cetool). All recombination products were transformed into DH5α cells, screened on medium containing 100 mg/L Ampicillin. Transformants were confirmed by plasmid DNA sequencing.

4.3. Protein expression and purification

The same protocol was applied for expression all proteins used in this study, including His₆‐MBP‐TEV‐MDPs, His₆‐MBP‐TEV‐Guy1, His₆‐GB1‐TEV‐Guy1, His₆‐TEV‐HSP90A, His₆‐TEV‐HSP90B, His₆‐GB1‐HSC70 and His₆‐TEV‐GB1. Briefly, E. coli BL21DE3 cells transformed with the plasmid coding for the corresponding protein were grown in 10 mL Luria‐Bertani (LB) media supplemented with 150 mg/L ampicillin at 37°C overnight. The preculture was incubated into 1 L LB medium in present of ampicillin (150 mg/L) to an absorbance at 600 nm (A₆₀₀ nm) range from 0.6 to 0.8. Protein expression was induced by adding 0.3 mM IPTG and cells followed to grow for 18 hr at 18°C. Cells were harvested by centrifugation at 5,000 g for 20 min at 4°C and resuspended in lysis buffer containing 25 mM Tris–HCl, 500 mM NaCl,10 mM imidazole, pH 8 (25 mM Tris–HCl, 1 M NaCl,10 mM imidazole and pH 7 for Guy1) with addition 1 mM PMSF and kept at −80°C for later use.

For protein purification, cells were lysed by a JN‐Mini Pro Low‐temperature Ultra‐high‐pressure cell disrupter (JNBIO) at 4°C. Cell lysate was clarified by centrifugation at 50,000 g for 30 min at 4°C. Supernatant was next passed through a Ni²⁺‐chelating column (GE Healthcare) pre‐equilibrated with lysis buffer. After sample loading, the column was washed with 25 mM Tris–HCl, pH 8.0; 500 mM NaCl (25 mM Tris–HCl, pH 7.0; 1 M NaCl and 25 mM imidazole for Guy1) to remove non‐specifically bound proteins. Protein eluted with 25 mM Tris–HCl, pH 8.0; 150 mM NaCl and 300 mM imidazole (25 mM Tris–HCl, pH 7.0, 150 mM NaCl and 300 mM imidazole for Guy1) from the column was then subjected to TEV protease cleavage for removal of amino‐end tags (His₆‐MBP, His₆‐GB1 or His₆‐tag) by dialysis, which was performed by putting dialysis bag containing elution protein and TEV protease miscible liquids to 50 mM Tris–HCl, pH 8.0, 100 mM NaCl and 1.5 mM βme overnight at 4°C. The protein mixture after TEV protease treatment was re‐loaded on to Ni²⁺‐chelating column to subtract amino‐terminal tag and His₆‐tagged TEV protease. The flow through was collected, concentrated and applied to a HiLoad 16/600 superdex 75 size exclusion column (GE Healthcare), equilibrated in 50 mM phosphate pH 7.0 and 100 mM NaCl; 3 mM βme (50 mM phosphate, pH 6.2, 100 mM NaCl and 3 mM βme for Guy1). The purity of each protein was assessed by SDS‐PAGE analysis and concentrations were determined by Epoch 2 Microplate Spectrophotometer (BioTek) at 280 nm using the corresponding extinction coefficient.

4.4. MALS experiments

MALS was measured using DAWN HELEOS II (Wyatt Technology Corporation) combined OptilabTrEX (Wyatt) system connected to a superdex200 increase10/300 GL (GE Healthcare) gel filtration column. The run buffer for all samples was 50 mM phosphate pH 7.0; 100 mM NaCl; 3 mM βme, except Guy1‐GB1 (50 mM phosphate pH 6.2; 100 mM NaCl and 3 mM βme). Protein samples at a concentration of 3 mg/mL were used. The flow rate was set to 0.5 mL/min with an injection volume of 100 μL and signal was collected at room temperature. The data were analyzed with ASTRA version7.0.1.

4.5. Circular dichroism measurements

CD spectra were collected on a Chirascan qCD spectrometer (Applied Photophysics Ltd.) with 0.1 cm path length quartz cuvette over a wavelength range of 195 nm‐280 nm at room temperature. Samples were prepared using phosphate buffer (50 mM phosphate pH 7.0; 100 mM NaCl; 3 mM βme) and the concentration of samples was 15 μm. The sample volume was 150 μL, time‐per‐point was 0.25 s, and bandwidth length was 1 nm. For each sample, three times repeat spectra collected were averaged, smoothed, and subtracted from the spectrum of three times average of GB1. The secondary structure was calculated by the CDNN software provided with the spectrometer.

4.6. Protein isotope labeling for NMR studies

Isotopically [¹H,¹⁵N]‐labeled samples were produced by growing the cells in minimal (M9) medium supplemented with 1 g/L of ¹⁵NH₄Cl and double‐labeled sample of humanin‐GB1 [¹⁵N,¹³C] were prepared for backbone assignment by adding 1 g/L ¹⁵NH₄Cl and 2 g/L ¹³C‐glucose in M9 medium.

4.7. NMR experiments

All NMR samples were prepared in 50 mM phosphate buffer, pH 7.0, 100 mM NaCl and 3 mM βme except Guy‐GB1 was prepared in 50 mM phosphate pH 6.2, 100 mM NaCl and 3 mM βme.

HSQC measurements were performed on a Bruker Avance 600 spectrometer equipped with a cryo‐probe with samples concentration of 100 μm and data were acquired at 383k. Backbone resonance assignments of humanin‐GB1 were achieved by using standard triple‐resonance experiments, measured at 383k with samples concentration of 330 μm in 50 mM phosphate buffer of pH 7.0 containing 100 mM NaCl and 3 mM βme. All NMR spectra were processed using NMRPipe program and analyzed using NMRViewJ software.

4.7.1. Accession IDs for proteins studied in this manuscript

Humanin: Q8IVG9; SHLP1: P0CJ68; SHLP2: P0CJ69; SHLP3: P0CJ70; SHLP4: P0CJ71; SHLP5: P0CJ72; SHLP6: P0CJ73; Guy1: J9ZV99, HSC70: P11142; HSP90α: P07900; and HSP90β: P08238.

AUTHOR CONTRIBUTIONS

Lin Huang: Conceptualization; data curation; formal analysis; investigation; methodology. Xiaozhan Qu: Conceptualization; data curation; investigation. Yao Chen: Data curation; investigation. Weiya Xu: Conceptualization; formal analysis; project administration; supervision; writing‐review and editing. Chengdong Huang: Conceptualization; formal analysis; investigation; project administration; writing‐original draft; writing‐review and editing.

Supporting information

Fig. S1: Comparison of SEC‐profiles of humanin‐GB1 purification using the sandwiched‐fusion strategy.

Fig. S2: {1H, 15N}‐HSQC NMR spectral overlay of GB1 with various peptide‐GB1 fusions.

Huang L, Qu X, Chen Y, Xu W, Huang C. Sandwiched‐fusion strategy facilitates recombinant production of small labile proteins. Protein Science. 2021;30:650–662. 10.1002/pro.4024