Native Chemical Ligation of Highly Hydrophobic Peptides in Ionic Liquid-Containing Media

The chemical synthesis of a highly hydrophobic membrane-associated peptide by native chemical ligation (NCL) in an ionic liquid (IL) [C2mim][OAc]/buffer mixture was achieved by employing peptide concentrations up to 11 mM. NCL was studied at different pH and water content and compared to several “gold-standard” ligation protocols. The optimized reaction protocol for the NCL in IL required the addition of 40% water and pH adjustment to 7.0–7.5, resulting in ligation yields of up to 80–95% within 1 to 4 h. This new ligation protocol is generally applicable and outperforms current “gold-standard” NCL methods.


■ INTRODUCTION
Ionic liquids (ILs) have been shown to be a good alternative to organic solvents for biotransformation reactions. 1−4 Several groups describe the use of ILs as advantageous solvents for protein refolding because of the ability of ILs with nucleophilic anions to either break or lower the formation of hydrogen bonds in solution, which leads to the suppression of aggregation in proteins. 5−8 The use of 1-methoxyethyl-3methylimidazolium hexafluorophosphate with a water content of 3% was described to be suitable for enzymatic peptide synthesis of tripeptide (ZTyrGlyGlyOEt), resulting in a higher enzyme activity and reaction yield in comparison to conventional systems. 9 Wehofsky et al. reported the use of 60% of 3methylimidazolium dimethylphosphate/buffer solution for protease catalyzed ligation. 10 The authors describe a good solubility of all reactants, the full suppression of proteolytic side reactions, higher turnover rates, and a higher stability of chemically unstable reactants. 10 Furthermore, 1-ethyl-3-methylimidazolium acetate ([C 2 mim][OAc]) was successfully used for native chemical ligation (NCL) and oxidative folding of various hydrophobic peptides. 11−14 In 2012, Boḧm et al. described the NCL in neat [C 2 mim][OAc] ligating a 30-and 36-mer to form the 66-meric polypeptide tridegin, a potent inhibitor of human blood coagulation factor XIIIa. 15 In neat [C 2 mim][OAc], the solubility of the peptide fragments was increased to >2 mM and the yield was significantly higher compared with that of the NCL in conventional ligation buffer. 15 In another study, the effect of the thioesterneighboring amino acid on the ligation yields in a neat [C 2 mim] [OAc], which contained about 3% water, was investigated. 16 Again, a faster and efficient ligation was achieved in comparison to conventional buffers. However, the yield of the ligation product decreased with time due to side reactions, such as oxidative folding or succinimide formation. 16 4 as a promoter to ligate a cysteine-free peptide fragment to a thioester peptide. 17 The ligation yields for di-, tri-, and tetrapeptides achieved by this method varied between 60 and 93%. Although ILs were used as reactants and solvents for biomolecules, their behavior and reaction mechanism differ in ILs and are not fully understood to date. Recently, we have made an attempt to understand the reactions of thiol-and disulfide-containing compounds in [C 2 mim][OAc]. 18 We have found conditions under which the IL can be used as a solvent or as a reactant through the presence of N-heterocyclic carbenes in the neat IL. 18 Based on these recent achievements, we sought to apply IL as reaction media for reactions which are problematic in conventional solvents, that is, conjugation of hydrophobic peptides or orthogonal modifications of peptides. Thus, the herein presented work will focus on the NCL of highly hydrophobic peptides in order to improve the synthetic accessibility to membrane proteins or fragments derived thereof, which in turn would allow for a more detailed structural and functional analysis of, for example, ligand-membrane protein interactions for development of efficient inhibitors of the viral influenza proton channel. 19−25 NCL is widely used as a standard method for the synthesis of soluble polypeptides, reaction conditions have been optimized, and the use of various thioesters has been introduced during the last decade. 15,26 However, NCL is challenging for highly hydrophobic peptides under standard conditions. 27−29 Several strategies were introduced to overcome the low solubility of hydrophobic peptides in buffer, for example, incorporation of removable or constantly attached solubilizing units 21 24 However, most of these strategies still suffer from low yields resulting from a poor efficiency of the ligation reaction and low solubility of highly hydrophobic fragments. Nevertheless, by increasing the NCL yield while eliminating solubility problems, we will compensate the product yield loss through aggregation during the sequence prolongation while synthesizing long fragments of hydrophobic peptides. 30,33 Herein, we describe an alternative and effective method for the NCL of membraneassociated peptides using [C 2 mim][OAc]. This method can be used as an alternative to removable solubilizing tag strategy 33 and does not require incorporation of additional solubilizing sequences into the core sequence of a hydrophobic peptide. For this purpose, we used fragments of different lengths of the influenza B proton channel (BM2) as a model system, which is a prototypical member of the proton channels of the influenza types A, B, and C and represents an important drug target. 37−39 First, model peptides were used to determine the best NCL conditions and finally applying an optimized NCL protocol for the NCL of the BM2(1−10) with the highly hydrophobic peptide fragment BM2 (11−51). Therefore, we determined the optimal reaction conditions while changing the ionic strength  . Rather than synthesizing thioester peptide fragments employing a sulfamylbutyryl resin, which usually requires a special workup procedure and gives rather low yields, we decided to use oxo-ester peptide fragments. This allows for the use of the more convenient rink amide resins, which only require a standard peptide workup and TFA cleavage. Moreover, the oxo-ester strategy allows for the incorporation of an in situ cleavable solubility tag, as recently shown by us. 33

■ RESULTS AND DISCUSSION
Thus, we first studied the NCL of two model peptides (1, 1′ and 2, Figure 1a) aiming to elucidate the impact of the ionic strength of the reaction solution and the yield (please note, the 2-hydroxy-3-mercaptopropionic acid (Hmp) oxo-ester was used as a racemic mixture resulting in two diastereomeric peptides named 1, 1′, Table 1 Table 1). In aqueous solutions, the only by-product that can be formed during this ligation reaction is the hydrolysis product of the oxo-ester fragment ( Figure 1b). As shown by Zheng et al., this type of oxo-ester undergoes a rearrangement through a 1,5-acyl migration with nearly 100% conversion, resulting in over 80% ligation yields. 40 All peptide fragments were successfully synthesized on a solid support using an AmphiSpheres RAM resin with 0.37 mmol/g loading capacity following a standard Fmoc-SPPS protocol, including coupling of the HMP group. The following amino acid was coupled through a Mitsunobu reaction. 41−43 Fmoc-deprotection after HMP coupling was performed with the mild base 2-methyl piperidine, to ensure stability of the HMP unit. Also, cleavage from the resin was performed following the standard cleavage protocol with 95% TFA as recently described by us. Thus, the synthesis of the thioetherforming fragment is straight forward to perform and does not require any special protocol, as is the case for sulfamylbutyryl resin-based thioester synthesis. Moreover, oxo-ester peptides are stable and can be stored for long time in comparison to thioester-peptides, which is a noticeable advantage of the selected strategy. Besides, we recently demonstrated that no epimerization of the amino acid at the condensation point was observed within a very similar strategy for the NCL of BM2(1−51), when oxoester peptides were employed. 33 In the present work, we studied the NCL of BM2 ( 40 The NCL reactions were followed by analytical RP-HPLC. The product formation was calculated with respect to the changes of the peak area of the product and the Cys-fragment. To avoid any bias in the data analysis resulting from significantly different extinction coefficients for the product and the Cys-fragment, we estimated the absorptivity of a reference solution for both compounds for the analytical HPLC setup. Thus, the peptide content of a standard solution of BM2 (22−35), BM2(17−35), BM2(11−51), and BM2(1− 51) was determined, and the peak area for each compound was recorded (Tables S2 and S4, more details are given in the Supporting Information). So far, within the error margin of the method, the absorptivity of the Cys-fragments BM2 (22−35) (2) and BM2(11−51) (6) and the respective ligation products BM2(17−35) (3) and BM2(1−51) (7) are similar (Table S4), allowing to directly compare their peak areas in the HPLC chromatograms.
First, we studied the influence of water on the ligation yields and the hydrolysis rates of the Hmp-group during the ligation   (Figures 2a and S5). All newly formed peaks in the RP-HPLC were separated and identified by electrospray ionization mass spectrometry (ESI-MS) ( Figures S3 and S4). The formation of the hydrolyzed product 4 ( Figure S3) and ligation product 3 ( Figure S4) was monitored for 6 h by RP-HPLC and their yields were calculated from relative peak areas (Figures 2a and S5).  Figure 2a) that at 50% water, the maximum amount of 60% of product 3 was formed within 4 h. However, after 4 h, the product began to precipitate under these conditions as indicated by the reduction of the peak area in the chromatograms. In contrast, when 40% water was present in the IL reaction mixture, only slightly less product 3 (53%) was formed within 6 h of reaction time and the product remained in the solution during the following hours (Figures 2a and S5). When the amount of water was reduced to 30%, the yield of product 3 was much smaller only reaching 25%. Thus, it seems that at an IL/water ratio of 60:40, the product formation appears optimal. Moreover, and contrary to our expectations, the relative peak area of the hydrolysis product 4 gradually increased with decreasing water content. In fact, this observation led us to analyze the pH of the IL/water mixture used for the ligation experiments. The pH values for the respective IL/water mixtures LB1a-f are outlined in Table S1. Please note that the read-out values for the pH were not corrected as suggested by Garcia-Mira et al. for water/GnHCl mixtures 44 since the specific correction factor for IL-based water mixtures has not yet been determined. Although measuring the pH of a neat IL is questionable, the pH of the IL/water mixtures decreased from 10.5 at a water content of 10% to 6.8 at 50% water. Apart from this pH dependency, the high viscosity of the IL/water mixtures when only low amounts of water were present might additionally reduce the reactivity of the ligation fragments. So far, these results are in agreement with previous results, which indicated an optimal pH of 7.5 for the ligation strategy used in this study. 33 In a second set of experiments, we aimed to keep the pH of the solution constant. Since acetate is the anion of the IL, concentrated acetic acid was used to adjust the pH, while increasing the amount of water and the ionic strength of the reaction solution. For this set of reaction mixtures (LB2a−f, Table S1), the amount of IL was in the range of 65 − 47%, while the amount of water was increased from 0 to 47%. With respect to this increasing water amount, glacial acetic acid was added to adjust the pH of the reaction mixture to 7.5. In contrast to the IL/water mixture of LB1, product 3 was already formed when no water was added to the initial mixture of IL and acetic acid (LB2a, Figures 2b and S6).
In contrast to the first set of experiments, the reaction progress was much slower for the first three sets of reaction (67/0/33, 64/7/29, and 64/16/20, Figure 2b), which contained either no or only a small amount of water (0, 7, and 16%, Figure 2b). Surprisingly, over the course of 6 h, a somewhat higher amount of product was formed (about 60 − 67%) when compared with the experiments (65/28/7 and 60/ 40/0.3, Figure 2b) with larger amounts of water (28, 40%, Figure 2b), which resulted in about 50 − 55% ligation product. Only 30% product was formed when the amount of water was further increased (47% at a reaction of 47/47/6, Figure 2b). Moreover, for this reaction product, precipitation was observed immediately after the reaction was started (Figure 2b). Interestingly, the progress and product formation of the ligation experiment in LB1e and LB2e (Figure 2a,b, 60/40 and 60/40/0.3) were almost identical, which can be rationalized by the almost identical pH and IL/water ratio (Table S1). Also, the overall amount of product formed during the ligation reaction was among the highest in these two IL-reaction mixtures. In an additional set of experiments, the water content was kept at 40%, which was suggested to be an optimal amount with respect to our data, while the reactions were performed at pH 7.0, 7.6, and 8.0 (Figure 2c, Table S1 and Figure S8). Whereas the reaction progress and product formation at pH 7.6 and 8.0 were similar, a somewhat slower reaction was observed at pH 7.0, but resulting in an almost quantitative product formation. Additionally, product 3 (BM2(17−35)) was isolated and purified by preparative HPLC from the reaction at pH 7.6 in this set of experiments, yielding about 35% purified product (95% purity). So far, the large amount of IL in the reaction mixture had no negative impact on the preparative scale HPLC purification. Usually, the IL fraction has no significant retention on standard C18 columns (250 × 8 mm and 250 × 20 mm were tested) and elutes shortly after the injection peak at low acetonitrile concentrations.
Lastly, a set of ligation experiments in conventional ligation buffer was performed using guanidine hydrochloride, urea, and a phosphate/SDS buffer system (LB3−LB5, and LB10, Figures  2d and S7). Because of the low solubility of the peptides 1, 1′ and 2 in ligation buffers LB3, LB4, and LB5, higher buffer volumes were necessary to entirely dissolve the peptides. Ligation experiments in these buffers were carried out at peptide concentrations of 1.3 mM (peptide 1, 1′) and 1.5 mM (peptide 2), which are 10 times more diluted compared to the IL-based buffers. The SDS-based phosphate buffer allowed for somewhat higher peptide concentrations of about 4 mM. The initial reaction rates of the NCL were higher in these standard buffers (Figure 2d) compared to the IL-based ligation solutions. The reaction was completed within 30 − 60 min and no further product was formed after that time, which agrees well with earlier reported reaction times. 40 In this set of ligation experiments, product precipitation occurred after some time in all reactions although much smaller peptide concentrations were used. Yields were about 20 − 25% The Journal of Organic Chemistry pubs.acs.org/joc Article lower in the conventional buffer systems, except for the NCL in GnHCl at pH 7.0 (Figure 2d), which has recently been shown to work exceptionally well for the NCL of BM2(1− 51). 45 Regarding the very fast reaction times in the standard buffer systems, choosing thiophenol instead of the better soluble MPAA as the ligation catalyst does not seem to have a significant impact on the reaction progress (Figure 3a). Moreover and as one would expect, more by-product 4 was formed, since aqueous buffer supports much faster hydrolysis of Hmp-peptide. Thus far, the results of these four sets of experiments clearly indicate that the water content and the pH of the IL-based reaction mixture have a strong impact on the reaction outcome with an optimal pH of about 7.0 − 7.5. The pH dependency, which was observed in all our experiments, can mainly be explained by the decreasing hydrolysis rate of the oxoester fragment when the pH of the reaction solution gets lower. At the same time, a certain amount of nucleophilic thiolate needs to be present, which usually requires a pH above 6.5 − 7.0. This behavior is in agreement with the standard reaction conditions necessary for an efficient NCL in standard solvents. 46 Furthermore, our data indicate no direct influence of the acetic acid on the reaction progress, except for the pH adjustment. As the addition of water had a stronger impact on the progress of the ligation reaction than acetic acid ( Figure  2a,b) and the viscosity of water and glacial acetic acid is similar, the viscosity of the reaction mixture alone is not the rate limiting factor.
Apart from this interpretation, it might be possible that the addition of water enables the formation of hydrated ion pairs of the IL components, which then are capable of better solubilizing the peptides or stabilizing reaction intermediates more efficiently than in the IL/acetic acid mixture without water (Figure 2b), thus resulting in an increased reaction rate. This would also be in line with recent results of Fujita et al., who found that specifically hydrated ILs are effective media for protein refolding from aggregates. 47 Taken together, our results show that IL-based buffers were clearly more suitable for ligation of hydrophobic peptides in comparison to standard ligation buffers under addition of chaotropic reagents. Although thiophenol appears to be not fully soluble at larger water contents (>40%) in the IL and in some of the buffer systems, which resulted in an increased suspension of thiophenol in the reaction mixture, the reaction progress does not indicate significant influence on the reaction performance ( Figure 2) which is most likely due to the large excess that was used. Also, with respect to the NCL reactions performed with the soluble thiophenol counterpart MPAA, which are outlined and discussed in the next section, no significant impact of the reaction performance could be observed ( Figure 3).
Last but not least, we compared the ligation protocol using an IL/water (60:40, LB1e) mixture with our recently developed and highly efficient 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP)-based ligation protocol (LB6) synthesizing a 50 aa long fragment of the BM2 proton channel 7 (BM2(1−51)), ligating peptides 5 BM2(1−10)-Hmp, and 6 BM2(11−51) (Figures 3a and S9−S12) employing MPAA as a ligation catalyst (please note: the Hmp-oxoester was used as a racemic mixture resulting in two diastereomeric peptides, namely, 5, 5, Table 1). The overall yield of the ligation in the HFIP-based reaction mixture (LB6) was slightly higher (∼10%) than that in the IL/water mixture (LB1e). Whereas the reaction in the IL/water mixture was almost completed after 4 h, the NCL in the HFIP-based solvent mixture was not finished even after 24 h (Figure 3a). Again, the IL/water mixture allowed us to use a 2-fold higher peptide concentration (2-fold) compared to the HFIP-based solvent mixture, although HFIP-based buffer systems were considered as more efficient than other standard buffer-based ligation solvent systems. 48 Thus, the higher concentration might partly explain the much faster reaction in the IL/water mixture (Figure 3a). With respect to the choice of the ligation catalyst, MPAA does not seem to perform significantly better, since similar reaction times were observed for the NCL of the short BM2 fragments, where it was used as the catalyst (Figures 2c and 3a).
Additionally, product 7 BM2(1−51) from the IL/water mixture at pH 7.5 was isolated and purified by preparative HPLC, resulting in an isolation yield of about 20%, which is half of the yield recently reported for BM2(1−51) employing a trityl-based solubility tag strategy in GnHCl/phosphate buffer. 45 The structural integrity of the BM2(1−51) peptide 7 synthesized in ligation buffer LB6 and LBe1 was confirmed by far-ultraviolet circular dichroism (CD) analysis (Figure 3b,c). So far, CD spectroscopic analysis yielded almost identical CD spectra for product 7 obtained from both ligation approaches and confirmed the α-helical nature of the BM2(1−51) fragment in TFE as well as after the fragment had been incorporated into a 1-palmitoyl-2-oleoylphosphatidylcholine (POPC) lipid. After deconvolution of the CD spectra, it was found that both BM2(1−51) peptides contain around 60% αhelical structure in TFE and in POPC.

■ CONCLUSIONS
In conclusion, the NCL of membrane-associated peptides by the reaction of a thioester-forming and a cysteine-containing peptide was successfully performed in IL [C 2 mim][OAc]. We were able to establish an efficient IL-based ligation buffer system which shows equal or even slightly better ligation yields compared to that of standard ligation buffers commonly used in modern peptide chemistry. The most efficient ligation buffer contained 60% IL and 40% water with an optimal pH between 7.0 and 7.5 and allows for much higher peptide concentrations in the ligation solution. Moreover, the IL-based ligation buffer was shown to be even more efficient for NCL of the highly hydrophobic membrane peptide BM2(1−51) in comparison to our previously developed NCL protocol in HFIP-based ligation buffer resulting in a faster product formation. CD measurement of BM2(1−51) confirmed the structural integrity of the BM2(1−51) fragment obtained from both ligation approaches.

■ EXPERIMENTAL SECTION
If not otherwise stated, all amino acid derivates and coupling reagents were purchased from carbolutions and used without further purification. Standard solvents for solid-phase-peptide synthesis and other standard chemicals were purchased from Sigma Aldrich and used without further purification. ILs were purchased from Iolitec ILs Technologies GmbH. pH of solutions were measured with either 716 DMS Titrino (Metrohm) or Five Easy F20 (Mettler Toledo). The pH values were used as a direct read-out from the pH meter and no correction was applied.
High-Performance Liquid Chromatography (RP-HPLC). All crude and purified peptides were analyzed by analytical RP-HPLC on a Waters 2695 Alliance system (Waters, Milford, MA, USA) employing a Waters 2998 photo diode array (PDA) detector equipped with a prontosil C8-SH (120 × 5 mm, 5.0 μm) column. HPLC eluent A was water [0.1% trifluoroacetic acid (TFA)] and eluent B was acetonitrile (0.1% TFA). If not stated otherwise, HPLC conditions for model peptides (1, 1′, and 2), 25−35% eluent B over 30 min at flow rate of 1 mL/min was applied (detection at 214 nm). For peptides (5, 5′, and 6), 10% eluent B for 3 min followed by 10 − 99% eluent B over 30 min was used. Chromatograms were extracted at 220 nm and analyzed at 214 nm. Peptide absorptivity for the 2998 PDA detector at 214 nm was determined from the peak area of a standard reference solution of the respective peptide. Concentration of the peptide solution was determined through a HPLC-based amino acid analysis protocol.
Preparative scale purification of the peptides was achieved by employing a Waters 1525 binary pump and a Waters 2998 PDA detector or a customized Waters 600 module equipped with a Waters 996 PDA detector (Waters, Milford, MA, USA). HPLC eluent A was water (0.1% TFA) and eluent B was acetonitrile (0.1% TFA).
Mass Spectrometry. The molecular weight of the purified peptides was confirmed by ESI mass spectrometry on a Bruker TOF-Q impact II spectrometer (Bruker Daltonik GmbH, Bremen, Germany) and calibrated using a Bruker's ESI-Tune-Mix or Waters SYNAPT G2-Si HD-MS spectrometer equipped with a Waters Acquity UPLC system (Waters, Milford, MA, USA).