Chemical synthesis of membrane proteins: a model study on the influenza virus B proton channel† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c8sc00004b

NCL results in the quantitative yield of a membrane protein, where a thioester peptide is formed from an oxo-ester with an in situ cleavable solubilizing tag.


Introduction
Membrane proteins are key drug targets as they are involved in essential processes in the cell, including the control of information and material ow between cells and the mediation/ propagation of nerve impulses. [1][2][3][4] The study of membrane proteins has led to new and improved pharmaceutical treatments for a wide range of illnesses such as migraine, multiple sclerosis, and cancer, as well as muscle and immune system disorders. Ion channels, such as proton, sodium (Na V ), potassium (K V ) or calcium (Ca V ) channels, are a type of membrane protein and are of immense importance in the human organism. 5,6 The molecular characterization of membrane proteins is essential in order to understand their functional properties, and to elucidate the mode of action of potent drug leads. Despite intense efforts to gather structural information on some classes of membrane protein as well as their drug binding properties, providing enough material for such studies remains challenging.
The synthetic production of highly hydrophobic peptides, such as medium-sized fragments of membrane proteins representing functional parts, offers many advantages over other production strategies, such as protein expression, mainly because synthetic peptides can be customized and derivatized depending on the purpose of the research. 7 Moreover, the production of multi-milligram amounts of membrane protein fragments by chemical synthesis might be even more important with respect to the development of bio-inspired materials. 8 The syntheses of highly hydrophobic peptides are usually performed using specically optimized synthetic strategies. 9,10 Peptides with more than 50 amino acids in length are usually obtained by fusing two or several peptide fragments through native chemical ligation (NCL). [11][12][13][14] The work up, including purication steps, for those peptides is usually challenging, since these peptides contain a high number of amino acids with hydrophobic side chains, which causes their aggregation in conventional solvents. 7,15 The rst attempt of the synthesis of inuenza A virus M2 membrane protein was performed by Kochendoerfer et al. The authors used a Boc-based SPPS, and for the ligation step they used 6 M guanidinium chloride containing 20% TFE in order to solubilize the peptide, resulting in 65% product formation. 12 Various methods have been established to overcome difficulties in the chemical synthesis of highly hydrophobic peptides, and they mainly focus on increasing the coupling efficiency 16 by improving the protocols of Fmoc-based solid phase peptide synthesis (SPPS). 17,18 The incorporation of pseudo-prolines was designed in order to reduce the aggregation during the chain prolongation by Fmoc-based SPPS, while at the same time being easily acid-cleavable. 19,20 Alternatively, the incorporation of depsipeptide or O-acyl isopeptide units has been used to increase the solubility of hydrophobic peptides. 20,21 However, even using these improved synthetic methods, purication remains extremely challenging. Aside from employing various mixtures of organic solvents 18,[22][23][24] and/ or C4, C1 or phenyl columns 18,24,25 for the HPLC purication of such peptides, the use of removable modications, [26][27][28][29][30][31][32] which can be incorporated to the peptide/protein's C-or N-terminus or backbone, was demonstrated to partly overcome the problem of handling hydrophobic peptides during their production. Therefore, positively charged amino acids, such as polyarginines or polylysines, were employed in order to increase the solubility of hydrophobic peptides. [33][34][35][36][37][38] However, these solubilizing tags are usually not removable and they alter the native sequence of the target peptide, which might impair peptide functionality. Consequently, removable solubilizing tags were developed which attach to the peptide via a cleavable linker (e.g. 4-hydroxymethyl benzoic acid (4-Hmb), 3,4-diaminobenzoic acid (Dbz) or 4-methoxy-5-nitrosalicylaldehyde), 30,39 an amino acid side chain/backbone, 28,29,40 or a short peptidase recognition site (which can be cleaved aer the acidic residues using carboxypeptidase B). 36,39,41,42 Nevertheless, only a few examples of cleavable solubilizing tags for chemical synthesis are described to date (Table S1 †).
Considering the NCL mechanism, the addition of solubilizing tags on the thioester-leaving group might be another promising strategy to increase the solubility of hydrophobic peptides, since the solubilizing tag is removed during the process of ligation, retaining the native peptide sequence. In 2007, Johnson et al. incorporated a polyarginine (Arg6) solubilizing tag attached to a thioester-leaving group using a Bocbased peptide synthesis protocol (Table S1 †). 33 Most of the examples described in the literature for the synthesis of membrane proteins and their parts were by Boc-SPPS (Table  S1 †), which usually requires special equipment, due to the use of HF in the nal global deprotection and cleavage from the resin used in SPPS. 43,44 Although, some Fmoc-based methods using peptide hydrazides were applied for membrane protein synthesis. 13,45 However, incorporation of a C-terminal solubilizing tag into peptide hydrazides is problematic. Therefore, a solution for this problem had to be found.
About a decade ago, an interesting Fmoc-SPPS compatible oxoester-based NCL strategy was described, which relied on the in situ generation of the thioester through an O-to-S acyl shi of the 2-hydroxy-3-mercaptopropionic acid (Hmp) moiety, which then reacts with the Cys-fragment to generate the nal ligation product. [46][47][48] Using this method, small glycopeptides, 47 the NNY-Rantes polypeptide chain (comprising residues 1-68), 49 and cyclotides 50 have been synthesized. However, under standard NCL conditions, high amounts (20-25%) of the hydrolyzed Cacarboxyester by-product were reported. 49,51 Moreover, the Hmp unit was found to be labile during standard Fmoc-SPPS, which might explain why this elegant strategy has not found a wide use. In 2013, Liu et al. observed that the use of 2-methylpiperidine instead of piperidine for the Fmoc-deprotection step nicely solved the instability problem during the Fmoc-SPPS of peptide-oxoesters. 51 At this point, we realized that this strategy can be employed to introduce solubilizing tags during Fmoc-SPPS that can be easily cleaved in a one-step reaction during the NCL procedure, resulting in a signicant improvement of the NCL of highly hydrophobic protein fragments. Consequently, the thioesterforming Hmp unit would serve as a cleavable linker between the solubilizing tag and the peptide chain.
Thus, our strategy combines the ligation of a conventional Cys-peptide fragment to an N-terminal fragment, which carries a removable, thioester-forming Hmp unit followed by a short PEG linker and a solubilizing tag. To demonstrate the feasibility of our anticipated strategy, we sought to synthesize small model peptides and an extended part of the membrane region of the inuenza B proton channel (BM2), which is a highly hydrophobic peptide and is prototypical for the class of small membrane-spanning ion channels. Moreover, BM2 represents an important drug target for the treatment of seasonal u, and its molecular structure has not been fully solved yet. [52][53][54][55] An efficient synthesis of such ion channel-forming peptides is of high importance, especially to the NMR community and their need for isotopically labelled samples.

Results and discussion
Development and optimization of the synthesis and NCL protocol for the Hmp unit using model peptides In order to optimize the synthetic strategy for the Hmp-peptide fragments (the thioester-forming peptide), a set of four different model peptides (1-4) was employed and ligated to a small Cyspeptide fragment (5) (Table S2 and S3, † and Fig. 1). The sequences of these peptides were derived from the inuenza virus BM2 protein sequence (Fig. 1a, Table S2 †), comprising residues 17 to 21 for the Hmp-peptide and residues 22 to 35 for the Cys-peptide fragment.
As described above, the Hmp unit was attached to a polylysine and/or mini-PEG solubilizing tag and incorporated into the model peptide fragments 1-4. More precisely, the thioesterforming peptides 2-4 were composed of four consecutive units: the peptide sequence, the Hmp unit, 49,51 a spacer (if necessary), and the solubilizing tag (Fig. 1b). Peptide 1 did not contain any solubilizing tag and/or spacer, and was used as a reference. During the NCL reaction, the Hmp unit forms the required intermediate thioester through an intermolecular S to O-acyl transfer, which is then followed by cleavage of the solubilizing tag upon formation of the MPAA thioester (Fig. 1b). Prior to the synthesis of the model peptides 1-4, the Hmp unit was synthesized as a racemic mixture according to the procedure described by K. Wisniewski. 56 The Hmp racemic mixture was directly used for Fmoc-SPPS, leading to a peptide mixture containing diastereomer A and diastereomer B which eluted in HPLC chromatograms at different retention times.
In the rst attempt, the native BM2 sequence comprising residues 17 to 21 (ALHFI, Fig. S3a, b †) was coupled to the Hmp unit, resulting in a very low ligation yield and a high degree of carboxyester hydrolysis (90%) (Fig. S1c †). This result was caused by the poor S to O-acyl shi yields when isoleucine (Ile21) is the C-terminal amino acid, as reported by Liu et al. 51 Therefore, Ile21 was mutated to Leu21, which did not show such a high degree of carboxyester hydrolysis. 51 Moreover, Ala22 was chosen as the respective ligation point, and was thus changed to Cys22 (Fig. 1a).
Polylysine (Lys 5 ) 57,58 or one or two 8-amino-(ADO) units were tested as solubilizing tags. 59 Additionally, ADO was used as a spacer in combination with the Lys 5 tag, in order to reduce any steric hindrance arising from the lysine Boc-protecting groups during the Hmp coupling (Table S2 †).
The thioester-forming peptides ALHFL-Hmp (1), ALHFL-Hmp-ADO (2), ALHFL-Hmp-ADO 2 (3), and ALHFL-Hmp-ADO-Lys 5 (4) were obtained as Hmp-diastereomers in yields between 20-60% (Table S2, Fig. S4, S5 †). The crude peptides 1-3 were directly subjected to the ligation experiments ( Fig. S4a-c †), whereas peptides 4-5 were rst puried by RP-HPLC (Fig. S4d, S5 †). Two dipeptides were synthesized in order to analyse whether the rst amino acid (L-Leu), which is coupled to Hmp via the Mitsunobu reaction, 60 is prone to racemization: L-Phe-L-Leu-Hmp and L-Phe-D-Leu-Hmp. HPLC analysis revealed different retention times for both dipeptides upon coelution, indicating that L-Leu does not racemize upon coupling to the racemic Hmp (Fig. S4e, f †). The desired peptides were stored at À20 C, and neither degradation nor hydrolysis was observed. This is a signicant advantage of our HMP-thioester strategy in comparison to methods where thioester-peptides, which are unstable at higher pH, are synthesized in a multistep reaction (i.e. on 4-sulfamylbutyryl resin). 61 The model peptides 1-4 were ligated to the Cys-fragment BM2(22-35) (5) at a molar ratio of 1 : 2 in ligation buffer A, and were very soluble under these conditions (Table S6 †). We noticed already at this step that peptide 4, possessing the Lys 5 solubilizing tag attached to the Hmp unit, is the least hydrophobic, eluting at $22% of eluent B (acetonitrile containing 0.1% TFA). This is in comparison to peptides 1-3, which all elute at $32% of eluent B, and possess ADO, ADO 2 or no solubilizing tag ( Fig. 2a-d). Consequently, ADO and ADO 2 did not inuence the solubility of the respective model peptide in comparison to the reference peptide 1.
The ligations for the model peptides 1-4 were completed within 2 hours (Fig. 2), and no further changes were observed at longer reaction times (>2 h), according to the RP-HPLC analysis. For peptide 1, the BM2(17-21)-MPAA thioester, which is generated through the in situ rearrangement of Hmp and reaction with MPAA, was already present directly aer starting the ligation. However, for peptides 2-4 the maximum amount of the thioester was formed at 30 min (Fig. 2b). The newly formed thioester-intermediate was subsequently converted into the MPAA-thioester, which was then ligated through trans-thioesterication with the C-terminal Cys of the BM2(22-35) (5) fragment. Aer the S to N-acyl shi, the nal peptide ALHFLCWTIGHLNQIKRGI (BM2(17-35) (6)) ( Fig. 1b and 2d) was obtained in 80-90% yield, according to the HPLC data (Table S3 †).
As already mentioned above and in accordance with the literature, the only byproduct formed during the ligation process was from hydrolysis of the carboxyester. 49,51 According to the RP-HPLC analysis, 13-16% of the side product ALHFL-CaOOH was formed during the ligation reactions of 1-4 with 5 (Table S3, † Fig. 2), which is comparable to the literature data. 49,51 However, the nal yields of the product 6 during the ligation reaction of the peptides 1-4 with 5 were similar, ranging from 80-90%. Hence, the introduction of a solubilizing tag into the model peptide 1, resulting in the peptides 2 (ADO), 3 (ADO 2 ) and 4 (ADO-Lys 5 ), did not impair the efficiency of the ligation reaction. Furthermore, our results indicate reaction times of less than 2 h for the NCL of 1-4 with 5.
Application of the NCL strategy for the synthesis of BM2  Encouraged by the results for the NCL of the model peptides 1-4 with 5, we decided to apply this strategy to longer and more hydrophobic peptides. Hence, we intended to synthesize a BM2 fragment which comprises residues 1-51, and now includes the full transmembrane domain and parts of the cytosolic part of the M2 protein. As for the model peptides 1-4, Ala22 was chosen as the respective ligation point, and thus mutated to Cys22  . Simultaneously, the cleavage of the solubilizing tag occurred. Then, the MPAA-catalysed transesterification was followed by an irreversible S to N-acyl shift. Finally, Cys22 was converted to Ala22 through desulfurization, and the Acm protecting group was cleaved from Cys11. (Fig. 1). Moreover, the thiol of Cys11 was protected with Acm in order to prevent a possible cyclization reaction (viz. reaction of the thiol group with the thioester) during the NCL. In analogy to the NCL strategy for the model peptides, ADO 2 and Lys 5 solubilizing tags were employed in combination with the Hmp unit (9-10, Table S2, Fig. S6 †) for the thioester-forming peptide [Cys 11 (Acm)]BM2 . Additionally, we synthesized BM2(1-21) (7) and [Cys 11 (Acm)]BM2(1-21)-Hmp (8) as references.
In analogy to the model peptides, the [Cys 11 (Acm)]BM2(1-21)-MPAA thioester was formed directly aer the reaction was started (Fig. 3), indicating that the MPAA-thioester formation does not depend on the peptide length. Interestingly, the ligation reaction took much longer to complete compared to the model peptides, almost fully converting the [Cys 11   and the thioester-forming Hmp unit, result in high yields (80%) of the desired product, and can be used generally for the NCL of highly hydrophobic peptide thioester-fragments. Even though our data already indicated a signicant improvement for the NCL of hydrophobic peptides, we sought to further improve the anticipated strategy by suppressing the carboxyester formation. Therefore, we used 2,2,2-triuoroethanol (TFE) or hexauoro-2propanol (HFIP), which were added to the ligation buffer A in a 1 : 2 (v/v) ratio, resulting in the ligation buffer B or C (Table  S6 †), respectively. TFE was previously described as an efficient additive of the ligation reaction for the synthesis of the M2 protein from inuenza A virus. 12,62 Ligation reaction in triuoroethanol/phosphate buffer (buffer B) In contrast to the NCL under standard conditions using the phosphate buffer (buffer A), the thioester-forming fragments 8-10 and Cys-fragment 11 were highly soluble in the TFEcontaining ligation buffer B. However, during the NCL of 9 and 11 (Fig. 4a), the product formation of [Cys 11 (Acm), Cys 22 ] BM2(1-51) (12) seemed to be completed aer 2 h, while the amount of the hydrolyzed carboxyester ([Cys 11 (Acm)]BM2(1-21)-C a OOH) continuously increased throughout the entire reaction process. Interestingly, the amount of the Cys-fragment [Cys 22 ] BM2(22-51) (11, t R 5.71 min) continuously decreased, with the continuous formation of an unknown side product (11 0 ) observed at t R 5.78 min (Fig. 4a). As concluded from the HPLC chromatograms, this side product appears to be unreactive towards the formation of the ligation product 12. A detailed analysis of this newly formed side product, using HPLC, ESI-MS and CD spectroscopy, revealed that 11 0 directly resulted from 11, presumably induced by TFE (Fig. 4).
This interpretation is further supported by the fact that 11 0 was directly formed in an aqueous TFE solution. In contrast, peptide 11 was totally stable in an aqueous HFIP solution, concluding that the strong helix-inducing effect of TFE and HFIP is not responsible for the formation of 11 0 , as further indicated by the CD spectra shown in Fig. 4d. Finally, MS spectrometry uncovered a difference of m/z ¼ 26 when comparing 11 0 to 11, indicating a TFE-induced chemical modication ( Fig. 4e and f), presumably with formation of a 2oxo-4-thiazolidine with the N-terminal Cys. 63,64 In 1987, D'Ischia et al. described the reaction of aminothiols with 1,1 0 -carbonyldiimidazole, resulting in the formation of 2-oxo-4thiazolidines. 63 The TFE-induced formation of a 2-oxo-4thiazolidine nicely explains the inactivation of 11 under the NCL conditions and the difference in M w of m/z ¼ 26. We suggest that urea would act in the same way as 1,1 0 -carbonyldiimidazole in this reaction, providing the carbonyl group at the 2 position of the thiazolidine ring. The calculated isotopic pattern for this suggested 2-oxo-4-thiazolidine modied [Cys 22 ] BM2  is in good agreement with the experimental isotopic pattern (Fig. 4f). Moreover, the formation of such a non-reactive side product nicely explained the rather high amount of the hydrolyzed carboxyester ($45%) under these conditions, since the Cys-fragment is not available for the ligation reaction, and the thioester fragment hydrolyses to a great extent. In particular, the reaction mechanism was not analyzed further, since we found that HFIP is a good alternative to TFE during the NCL, without forming any unreactive product with the peptide.

Ligation reaction in hexauoro-2-propanol/phosphate buffer (buffer C)
Because of the stability of 11 in HFIP (Fig. 4c), a third ligation buffer system (Table S6 †) was tested. Due to the reduced solubility in HFIP of the ligation additives TCEP, MPAA and urea, they were used in slightly lower concentrations ( Fig. 4 and 5). As outlined in Fig. 5, the NCL of the peptide fragments 8-10 with 11 was completed within 24 h, almost quantitatively forming the product 12a. For [Cys 11 (Acm)]BM2(1-21)-Hmp (8) and [Cys 11 (Acm)]BM2(1-21)-Hmp-ADO 2 (9), we were able to separate the two diastereomers which resulted from the racemic HMP building block used for the synthesis of 8-10, allowing us to conclude that they are equally reactive. They also form similar and low amounts ($4-12%, Table S4 †) of hydrolyzed carboxyester. The reason for the lower hydrolysis rate observed in the ligation buffer C might be due to a lower water content in the buffer C in comparison to the buffer A, thus resulting in a faster reaction with the Cys-peptide.

Desulfurization and Acm-group deprotection of BM2(1-51) (12a)
Aer the successful ligation of the BM2(1-51) proton channel through our HMP-strategy we focused on the backtransformation of the modications, which we introduced with respect to our ligation strategy, namely cysteine at position 21 and an Acm protecting group at position 11. Hence, Cys21 was transformed into Ala through a desulfurization step, followed by cleavage of the Acm protecting group from Cys11.
Desulfurization of the Cys residues expands the number of possible ligation sites. 65 This is especially of importance for peptides which have no naturally occurring or suitable Cys residues for the anticipated ligation strategy.
The desulfurization of Cys21 was achieved using the method introduced by Danishefsky et al., 66 which was optimized due to the insolubility of peptide 12a in the desulfurization buffer.
A nearly quantitative conversion of Cys to Ala was achieved for the peptide 12a (within 6 h, Fig. 6b, Table S5 †), yielding peptide 12b. Interestingly, we found that by using this method, almost quantitative desulfurization can be accomplished for the soluble peptide 5 within 4 h (Fig. S7 †). This result is similar to that described for small hydrophilic peptides in the literature, however the yield was reported to vary (80-99%) depending on the peptide sequence. 65,66 Finally, for the cleavage of the Acm protection group from Cys11 to obtain 12c, three different methods were tested using iodine oxidation, 67,68 a silver salt, [69][70][71][72][73][74] and DTNP. [75][76][77][78] The DTNP approach, however, failed in our hands, probably because of the presence of Met and Trp in the primary structure of our peptide 12. 79 Using iodine oxidation we achieved only very poor yields ($16%) of the nal product 12c (Fig. S8 †). The Acm deprotection employing silver triuoromethanesulfonate (AgOTf) resulted in a nal yield of $60% of 12c (Fig. 6c) aer HPLC purication, which is comparable to the literature results for the hydrophilic peptide oxytocin. 70 The peptide 12c was nally puried (Fig. 6d and e, Table S5 †) and subjected to CD analysis.

Structural indications of the synthetic BM2(1-51)
In order to analyze whether the BM2(1-51) protein fragment (12c) would correctly fold, the nal product 12c was subjected to CD analysis. Additionally, the far UV CD spectra of BM2(1-21) (7) and BM2(1-51) (12a) at 20 C were rst measured in TFE (Fig. 7). All spectra showed the characteristic double minima at 209 nm and 223 nm, which are typical for an a-helical fold (Fig. 7). Aer deconvolution of the CD spectra, it was found that the peptides 12a and 12c and fragment 7 largely exist in an ahelical form (Table S7 †), containing around 70% a-helix. As a reference, the BM2(1-33) fragment revealed around 60% a- helical content under these conditions. Moreover, deconvolution of the CD spectra of the BM2 fragments 7 and 12c and BM2  in POPC (1-palmitoyl-2-oleoylphosphatidylcholine) revealed a similar a-helical content of approx. 60% for all three BM2 fragments.
These ndings are also in line with the recent structural data from a solid-state NMR study of BM2(1-33), which indicated an almost fully a-helical structure. 54 Consequently, BM2(1-51) (12c) appears to be structurally very similar compared to its shorter counterpart BM2 , and this might indicate some ordered helical structure for the region between residues 34 and 44, which was predicted to be structurally disordered from a solution NMR study of the full length BM2. 54

Conclusion
In conclusion, the use of a removable solubilizing tag combined with an oxo-ester functional unit is an effective method for the synthesis and purication of highly hydrophobic peptide fragments. The oxo-ester peptide can be effectively used for the native chemical ligation with cysteine-containing peptide fragments, as demonstrated by the synthesis of the inuenza virus B proton channel fragments BM2 (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35) and BM2(1-51). Moreover, our strategy allows the incorporation of various solubilizing tags into an oxo-peptide sequence, which was shown by the example of ADO, ADO 2 or ADO-Lys 5 tags. For the native chemical ligation, the Ala22 residue was replaced with Cys22, and Cys11 was protected with an Acm-group. Both modications were fully reversible through desulfurization (99% yield) and Acm-group cleavage (60% yield) aer the careful optimization of existing literature protocols for our hydrophobic peptides. Finally, circular dichroism analysis of the BM2 fragments, BM2(1-21) and BM2(1-51), in TFE and reconstituted in POPC revealed that the synthetic peptide exists mainly as an ahelical structure, similar to the recently characterized BM2 , indicating that the structurally disordered region between residues 34-44 most likely possesses an a-helix structure.
The current study will provide an important alternative method to prepare highly hydrophobic peptides and proteins, allowing their synthesis in a multi-milligram scale, i.e. for future structural and binding studies.

Synthesis of 3-chloro-2-hydroxypropanoic acid
A 3 M solution of 3-chloro-1,2-propanediol (0.09 mol) in 30 mL nitric acid (conc.) was slowly heated to 80 C until a vigorous reaction started. The temperature was kept at 80 C for 20 min and then warmed-up to 100 C for 30 min. Subsequently, the reaction solution was cooled to room temperature and neutralized with sodium bicarbonate. The product was extracted with diethyl ether, and the organic layer was dried over anhydrous sodium sulfate before the solvent was removed under reduced pressure. The residual solvent was evaporated to leave a viscous liquid. The product was precipitated in cold chloroform and ltered to obtain 7.4 g of 3-chloro-2hydroxypropanoic acid (yield: 66%). NMR (DMSO-d 6 ): d 3.77 (2H, CH b ), 4.30 (1H, CH a ) (Fig. S1 †).  Synthesis of 2-hydroxy-3-(triphenylmethyl)thio-propanoic acid (Hmp(Trt)-OH) A 1 M solution of 3-chloro-2-hydroxypropanoic acid (0.047 mol) in dry 1,2-dimethoxyethane was cooled to 0 C, and sodium hydride (60% m/m dispersion in mineral oil, 0.047 mol) was added slowly. A solution of triphenylmethanethiol (0.05 mol) and sodium hydride (60% dispersion in mineral oil, 0.047 mol) in dry 1,2-dimethoxyethane was added dropwise, and the resulting reaction mixture was stirred for 4 h. Aerwards, the solvent was removed under reduced pressure yielding a yellow, solid residue, which was dissolved in a diethyl ether/water mixture (1 : 1). The aqueous phase was separated, washed with diethyl ether and acidied with 1 M hydrochloric acid. The product was extracted with ethyl acetate. Finally, the organic layer was dried over anhydrous sodium sulfate and the solvent was removed under reduced pressure until a highly viscous liquid was le. The liquid was solidied overnight in the refrigerator and the rest of the solvent was removed under reduced pressure until a slightly yellow powder was obtained. Yield: 11.08 g (98%). 1 (Fig. S2 †).
The Hmp-containing model peptides (1)(2)(3)(4) were synthesized manually on the resin. For the extended Hmp-containing peptides (8)(9)(10): the solubilizing tag (Lys 5 , ADO 2 ), spacer (ADO), Hmp unit and the rst two amino acids of the respective peptide fragment were coupled manually to the resin. The following amino acids were coupled (30 min, double coupling) using an automated peptide synthesizer (Liberty, CEM) applying identical conditions as those for peptides 5 and 11. The S-acetamidomethyl (Acm) group was used as a side chain protection group for Cys11. For the peptides 8-10, Fmoc-groups were removed with 20% 2-methylpiperidine (v/v) in DMF.
The amino acid Leu21 (Ile21) was coupled to the preceding Hmp-unit as described by Liu et al. 51 Accordingly, the resin was re-swollen in dry oxolane (THF) and treated twice (double coupling) with a 0.2 M mixture of the amino acid (4 equiv.), triphenylphosphine (3.9 equiv.) and diethyl azodicarboxylate (40 wt% in toluene, 3.9 equiv.) for 2 h. Aerwards, the resin was washed intensively with THF.
Peptide cleavage and deprotection was accomplished in a mixture of 90% TFA, 5% water, 2.5% TIPS and 2.5% anisole for 3 h at room temperature. The crude peptides were precipitated in cold diethyl ether, centrifuged, and washed with diethyl ether.

Native chemical ligation
The ligation experiments were performed in the ligation buffers (Buffer A, B and C, see Table S6 †) at a 1 : 1.5 (n/n) ratio of the thioester-forming peptide (1-4, 8-10, 1 mM) and Cys-peptide (5, 11, 1.5 mM). The ligation reaction was initiated by adjusting the pH of the ligation buffer to pH 7.0 or pH 7.5 with NaOH (1 M) depending on the used buffer (ESI Table S6 †). All ligation experiments were performed under an N 2 atmosphere.

Desulfurization of [Cys 11 (Acm), Cys 22 ]BM2(1-51)
A 0.5 M tris(2-carboxyethyl)phosphine (TCEP) HCl solution in 6 M guanidinium hydrochloride (GdnHCl)/0.2 M disodium phosphate buffer was prepared. The pH of this solution was adjusted to 6.5 and reduced glutathione (GSH, 125 mL of a 160 mM solution in water) was added. The reaction was performed under a nitrogen atmosphere. 1.2 mg of the peptide [Cys 11 (Acm), Cys 22 ]BM2(1-51) was dissolved in 2 mL HFIP and added to the reaction solution. Aerwards, 200 mL of a VA-044 solution (300 mM in water) was added and le to react for 6 h. The reaction progress was monitored by analytical RP-HPLC (C4 column 125 Â 4 mm, 300Å pore diameter, and 5.0 mm particle size) with a linear gradient of 45-70% eluent B over 30 min at a ow rate of 1 mL min À1 . The reaction was stopped by transferring the lower layer into the 4-fold volume of 0.1% triuoroacetic acid (TFA). The solution was freeze-dried and the peptide was isolated by preparative RP-HPLC using a C4 column (MultoHigh Bio 300-5, 250 Â 20 mm, 300Å pore diameter, and 5.0 mm particle size) with a linear gradient of 40-80% eluent B over 50 min at a ow rate of 10 mL min À1 . Yield: 1.09 mg [Cys 11 (Acm)]BM2(1-51) (91%).

Conflicts of interest
There are no conicts to declare.