Intein-mediated recombinant expression of monomeric B22Asp desB30 insulin

Background Insulin controls hyperglycemia caused by diabetes, and virtually all treatments require exogenous insulin. However, the product’s extensive post-translational modifications have hindered the manufacture of recombinant insulin. Result Here we report a novel production method for a monomeric B22Asp desB30 insulin analog (B22D desB30 insulin). Its precursor, DPIP, is fused to an N-terminal chitin-binding domain and intein self-cleavage tag. The fusion protein is expressed and purified from E. coli and immobilized on chitin resins. DPIP is then released using an optimized pH shift and converted to mature insulin via trypsin digest. The resulting product appears monomeric, > 90% pure and devoid of any exogenous enzyme. Conclusion Thus, biologically active insulin analog can be efficiently produced in bacteria and potentially applicable in the treatment of human diabetes.


Background
Diabetes mellitus in human causes elevated blood sugar levels for a prolonged period of time [1]. If untreated, the disease may progress into many life-threatening complications, like kidney disease, blindness, and amputations [2]. Virtually all patients require regular intake of exogenous insulin, which lowers their blood glucose concentration. Studies suggest that the disease affects hundreds of millions of people, with an annual cost rapidly approaching one trillion dollars [3].
Since 1977, production of animal insulin has gradually exited the market as recombinant human insulin emerged [4]. However, since the human insulin monomers readily aggregate into multimers, the recombinant wild type human insulin displays a delay in treating hyperglycemia [5]. Therefore, its monomeric analogs, termed fast-acting insulins, are developed and clinically proven to match human endogenously produced insulin [4,6]. However, both the manufacturing process and composition of these patented insulin analogs are closely guarded commercial secrets. Due to the increasing diabetic population and rapidly rising cost, the market urgently requires the development of low-cost, fast-acting insulin analogs, especially in the underdeveloped and developing nations.
Here we report recombinant production of a fastacting insulin analog, termed B22D desB30, matured from a single-chain precursor. In vivo assays have shown that analogs of B22D desB30 stays monomeric and exhibits 30-40% activity, compared to the human endogenous counterpart [7][8][9]. However, structural changes that prevented multimerization also possibly destabilized its overall folding and hence made production difficult [10]. Therefore, we opted to recombinantly express and purify the B22D desB30 precursor by protein splicing, which yielded single chain insulin precursor (B22D-PIP or DPIP).
Protein splicing involves the precise excision of an intein from a primary translation product concomitant with the ligation of the exteins via a peptide bond [11]. This autocatalytic process occurs post-translationally, without other enzymes or any cofactors [12,13]. Several biotechnological applications explore the splicing properties of inteins, including protein purification, peptide cyclization and protein labeling [14][15][16]. Among them, the intein mediated purification with an affinity chitin-binding tag (two Intein or IMPACT-TWIN by New England Biolabs) provides a low-cost, convenient system, in which the target protein is immobilized by a single-step affinity enrichment and collected by intein self-cleavage that removes the affinity tag [17,18]. Using this process, we obtained the native protein without any affinity tag or exogenous proteases. Subsequent tryptic digestion produced pure, monomeric B22D des30 insulin analog.

CBD-intein1-DPIP fusion protein expresses in the inclusion body
The recombinant B22D desB30 produced in this study is converted from a precursor (DPIP) fused to a CBDintein tag. Fig. 1a shows the recombinant expression scheme of B22D desB30 insulin analog. The construct CBD-Intein-DPIP was confirmed by sequencing and all expression was carried out at similar conditions. The recombinant fusion protein is consistently expressed at a high level (Fig. 1b, lane 2). Most likely the target protein collects in the inclusion body and is insoluble. The fusion protein expresses well in the inclusion bodies and was insoluble, consistent with precious studies [20,21]. After collecting the pellet after cell lysis, the fraction that redissolved after urea treatment is enriched with the 36 kDa target protein (Fig. 1b, lane 4). Solubilizing the inclusion bodies with urea allows the ready purification of the fusion protein to near homogeneity. Subsequent affinity chromatography traps nearly all renatured fusion protein on the chitin matrix. All renatured protein binds to the column after dialysis, which suggests that the re-naturation process correctly restored native protein folding (Fig. 1b, lane 5). The CBD-intein tag also allows a pH-induced auto-cleavage at the Nterminus of the insulin precursor, releasing the peptide without exogenous protease or tags.
Since intein cleavage is sensitive to both temperature and pH, we aim to fine-tune the reaction condition to maximize purification efficiency. This allows the most efficient release of the target DPIP with simple temperature and pH shift. By carrying out the cleavage reaction at three different temperatures (10°C, 25°C and 30°C) and eight pH conditions (4.5 to 8.0, 0.5 interval), our study shows that the intein selfcleave reaction is strictly temperature-and pH-dependent. During optimization of the pH-induced intein selfcleavage, we discovered that lowering the pH below physiological conditions aided the cleavage process. Fig. 2a shows proteins retained by the chitin column by SDS-PAGE after reaction at various temperature and pH. Lane 1 shows that the fusion protein (CI-DPIP, 36 kDa, marked by the top arrow) bound to the column before cleavage. Lanes 2 through 8 indicates that one of the products (CBD + intein1, 31 kDa, marked by the lower arrow), which also tightly associates with the column, decreases in intensities with the increase of pH (quantified in Fig. 2b). In general, we observed an increase in reaction activity (almost 100%) at lower pHs compared to physiological pH. This result is somewhat unexpected given the proposed mechanism for Cterminal intein cleavage involves deprotonation of the amino group [22].
This observation coincides with previous results [19,23] in which the intein C-terminal cleavage rate increased when the pH was lowered to 6.0. Therefore, we speculate that this phenomenon maybe general, and acidic conditions may benefit the self-cleavage reaction of other intein fusion proteins. Our data suggest that this particular Cterminal intein cleavage is perhaps catalyzed by a general acid (hydronium) via alternative mechanisms. However, at extremely acidic environments (pH < 5.5), we suspect that DPIP (pI = 4.9) may become unstable or insoluble. Therefore, we determine that the optimal cleavage pH for CI-DPIP is 5.5. Previous studies suggest that intein may undergo different reaction mechanisms at varying pHs. However, no previous study has investigated intein cleavage conditions below pH 6.0 either experimentally or theoretically. Theoretical approaches proposed that at lower pH (6.0), especially below the pKa of histidine (6.04), the C-terminal cleavage reaction may take a different reaction path and increase its reaction rate [24]. Their result suggests that N-protonation of the scissile peptide bond likely starts the C-terminal cleavage reaction, in contrast to a nucleophilic side chain cyclization mechanism proposed at physiological pH [22]. Our results support the acidcatalyzed reaction mechanism hypothesis at low pHs.
While having very little effect on cleavage efficiency at higher pHs, temperature significantly increases the reaction rate at pH 6.5 and lower. Among all three temperature conditions tested, 25°C shows, on average, 40% higher activity than 10°C and 30°C at pH 6.5-4.5. Although higher temperatures usually produce higher reaction rates, CI-DPIP cleaved inefficiently at 30°C, similar to 10°C. This result suggests that the intein structure in this construct may be unstable and the active site perturbed at 30°C. Our above data suggest that a simple pH change at 25°C can induce the release of tag-less DPIP from the recombinant protein. At pH 7.5-8.0, the cleavage rate is the slowest (below 5%), which means that the precursor can be stored safely under these conditions. Once shifted to 25°C and pH 5.5, the intein self-cleavage reaction rate constant increased significantly to 30%. Therefore, we conclude that the optimal temperature and pH for CI-DPIP cleavage in this study is 25°C and pH 5.5. By shifting the buffer pH from 8.0 to 5.5, we can release pure insulin analog precursor DPIP from the column.

Intein cleavage increases over two days
After determining the optimal temperature and pH for the intein cleavage reaction, we set to improve the yield of DPIP by investigating different reaction durations. We had explored reaction time on the self-cleavage of intein by comparing the resin samples after 2, 4 and 6 days of cleavage. During a six-day incubation, resin samples were collected separately every other day. Subsequent SDS-PAGE (Fig. 3a) and band analysis (Fig. 3b) showed that the reaction rate increases drastically over the first 2 days and leveled off after that. Although small amounts of product continued to form over the next four days, the amount of additional product cannot justify the extra time required and the risk of degradation. Therefore, two days was chosen as the optimal reaction time.

Urea inhibits intein self-cleavage
Our study shows that the CI-DPIP construct is sensitive to urea. Since previous findings show that low concentrations of urea may increase the intein cleavage ratio [20], we decided to investigate the effect of urea on CI-DPIP self-cleavage. We tested two urea cleavage conditions (1 M and 2 M) against the no urea TGE buffer, and the results show that even one-molar urea can reduce reaction efficiency by 43% (Fig. 4). At two-molar urea concentration, the reaction efficiency dropped to below 10%. This result suggests that our CI-DPIP fusion protein is more sensitive to folding, and slight perturbation abolishes the intein activity. This is also consistent with our temperature study that higher temperatures adversely affect the cleavage reaction, presumably due to misfolding at higher temperatures. It is possible that Fig. 3 Rate of CI-DPIP cleavage increases then slows down after two days. a: SDS-PAGE analysis of cleavage with different incubation time. Lanes 1-3 indicate resin after incubation of 6 days, 4 days and 2 days, respectively, and lane 4 indicates the input precursor without any cleavage (0d). Since the precursor is chemically stable in solution, so Lane 4 shows no CBD + intein fraction, M: molecular weight markers. b: Quantified cleavage ratio. All assays were carried out in triplicate and the average was used in this study; the error bars show standard deviations urea and higher temperate adversely affect CI-DPIP splicing by similar mechanisms.

DPIP precursor generates pure B22D desB30 insulin analog
Since the smaller size of DPIP (MW = 5.7 kDa) does not allow accurate analysis on SDS-PAGE (Fig. 5a), 16.5% tricine gels were required to visualize and analyze the fragment (Fig. 5b). Our result clearly identifies the pure DPIP eluted from the chitin matrix using conditions described above (Table 1).
To obtain the final folded insulin monomer, DPIP was subjected to trypsin digest. This step releases A-chain and B-chain from DPIP, removing the linker peptides (GRAK at the N-terminus and AAK between A-chain and B-chain) (Fig. 1a). Subsequently, two disulfide bonds spontaneously link A-chain and B-chain to form the final monomeric product. This insulin analog migrated as a single peak at 19.5 min during our HPLC purification (Fig. 6a), and subsequent mass-spectroscopy (Fig.  6b) confirmed a molecular mass of 5664 (theoretical MW = 5663). These data confirm that we have obtained pure (> 90%) and correctly folded B22D desB30 insulin analog through recombinant expression and purification. Table 2 summarizes total yield and purity of our DPIP purification.
We further determined the quality and monomeric state of the B22D desB30 insulin by comparing our sample with commercially available monomeric recombinant insulin through size exclusion chromatography (Fig. 7). The recombinant B22D desB30 insulin eluted as a single peak at 15 min, which coincides with the major peak in the commercial sample, confirming the purity and monomeric nature of the B22D desB30 analog. Moreover, our sample showed no detectable degradation or fragmentation (Fig. 7b). Since the target peptide remains  stable, we recommend using large-scale SEC to separate it for mass production.

Conclusion
This study aims to explore a cheaper alternative for producing monomeric human insulin analog to treat diabetes. Compared to the chemical synthesis and yeast preparation, our recombinant monomeric B22Asp desB30 insulin analog is fused to an N-terminal chitinbinding domain and intein self-cleavage tag. The fusion protein binds to chitin resins and is subsequently washed off. Next, trypsin digest releases mature insulin.
The resulting product appears monomeric, > 90% pure and devoid of any exogenous enzyme. Thus, biologically active insulin analog can be efficiently produced in bacteria and potentially applicable in the treatment of human diabetes.

Methods
All the buffers used in this study are summarized in Table 2.

Protein expression
All expression experiments were performed in duplicate with good reproducibility. First, pTWIN1-DPIP were transformed into E. coli strain BL21 (DE3) and grown to optical density (A600) of 0.5, in 300 ml of LB media (2% tryptone, 1% yeast extract, 2% NaCl, w/v) containing 100 μg/ml ampicillin at 37°C. Next, isopropyl 1-thio-β-D-galactopyranoside (IPTG) was added to the final concentration of 0.1 mM to induce expression for 4 h at 37°C. The cells were harvested by centrifugation at 3000 g for 15 min at 4°C and stored at − 80°C.

Protein purification
All subsequent steps in this section were performed on ice or at 4°C unless otherwise mentioned. Cell pellets were resuspended in ice-cold PBS (pH 8.0) at a ratio of 1:10 (w/ v) and lysed by sonication. Since the DPIP is enriched in the insoluble fraction, the cell extracts were centrifuged at 10,000 g for 30 min at 4°C to remove the soluble fraction. The pellets were washed sequentially with buffer A-E before redissolved thoroughly in ice-cold buffer F. Next, the mixture was dialyzed against TGE buffer (pH_8.0) with different concentrations of urea (6 M, 4 M, 2 M, 1 M, and 0 M). The insoluble fraction was removed by centrifugation at 10,000 g for 30 min.

Quantifying the intein splice efficiency
Premature cleavage was estimated by quantifying the scanned SDS polyacrylamide gels with Bandscan 5.0 (Glyko). Bands corresponding to full-length fusion protein (CBD-intein1-DPIP, CI-DPIP) and CBD-intein1 tag (CI) were quantified by scanning densitometry and normalized against their molecular weights.

Optimizing tryptic digestion
The eluate from intein splicing was digested by trypsin at 30°C for one hour to obtain the final B22D desB30 insulin analog. We used an enzyme/substrate ratio of 1: 200 (w/w). The mixture was subsequently purified by high-pressure liquid chromatography (Wasters) (C8 column, acetonitrile and trifluoroacetic acid (TFA) as mobile phases). Monomeric B22D insulin analog was eluted at 19.5-min mark with a flow rate of 1 ml/min and a buffer H gradient of 35-50% over 30 min.
To access the quality of our sample, both the commercial insulin (Insulin Injection, MW 5778) and the B22D desB30 insulin analog was subjected to size exclusion chromatography (SEC) using Superdex 75 (GE healthcare) column under standard conditions.