Total synthesis of interleukin-2 via a tunable backbone modification strategy

Chemical synthesis of hydrophobic proteins presents a formidable task as they are often difficultly achieved via peptide synthesis, purification, and peptide ligation. Thus, peptide solubilizing strategies are needed to integrate with peptide ligation to achieve protein total synthesis. Herein, we report a tunable backbone modification strategy, taking advantage of the tunable stability of the Cys/Pen ligation intermediate, which allows for readily introducing a solubilizing tag for both peptide purification and ligation processes. The effectiveness of this strategy was demonstrated by the chemical synthesis of interleukin-2.


Introduction
The past two decades have witnessed the rapid development of innovative peptide ligation methodologies 1-9 and strategies, [10][11][12][13] which have led to dramatic changes in the scope of protein chemical synthesis. Many proteins of biological interests have been chemically synthesized and used to correlate the protein structure to function. [14][15][16][17][18][19][20][21][22][23][24] In particular, protein chemical synthesis provides a unique means to generate proteins with tailor-made modications, which are difficult or impossible for expression systems. [25][26][27][28][29][30][31][32][33][34][35][36] Despite these achievements in protein chemical synthesis, the synthesis of proteins or peptides with aggregation-prone properties remains a challenging task and requires case by case analysis and study. The dilemma can be classied into two types: (a) the peptides aggregate on the resin beads during solid-phase peptide synthesis (SPPS), leading to peptide elongation failure or poor product quality; 37,38 (b) the peptide products tend to form insoluble or colloidal solids aer successful synthesis and cleavage from the resin, which makes reverse-phase high-performance liquid chromatography (RP-HPLC) purication or performing peptide ligation difficult. 39 To address this difficulty, many efforts have been devoted, involving the employment of solubilizing tags, [40][41][42][43][44][45][46][47][48][49][50][51] O-acyl isopeptide, 52-56 removable backbone modication, 57-59 and special ligation solvent conditions. [60][61][62][63][64] Interleukin-2 (IL-2) is a cytokine playing an important role in cancer immunotherapy. Indeed, in clinical treatment, the effect of IL-2 is a bit complex because of its dual functional roles in T cell. 65 It can not only promote activation and proliferation of natural killer cells and cytotoxic T cells to destroy tumor cells 66 but also enhance regulatory T cell activities to downregulate T cell cytotoxicity. 67 Therefore, the chemical synthesis of IL-2 could offer an opportunity to construct site-specic modied IL-2 variants with more focused biofunctions, which may lead to better clinical outcomes in cancer or immunosuppression treatments.
IL-2 is a typical difficult protein for chemical synthesis. To achieve the goal of chemical synthesis of IL-2, the primary task is to overcome the obstacle of preparation of the extremely hydrophobic and highly aggregated C-terminal region. In 2015, aer more than 15 years of efforts, Hojo's group achieved the rst total synthesis of IL-2 using both O-acyl isopeptides and picolyl ester protecting groups. 54 Later, Bode's group reported the second synthesis of IL-2 by performing KAHA ligation at several mutated sites and using O-acyl isopeptides for segment preparation. 55 The two studies showcase the current state-ofthe-art of protein synthesis. However, new NCL-compatible strategy to enable the robust synthesis of the IL-2 C-terminal region is appealing but remains to be explored, which will expand the NCL application scope to such type of highly hydrophobic and aggregated peptides.
Herein, we report a tunable backbone modication (TBM) strategy ( Fig. 1) to enable the robust and straightforward synthesis of IL-2. Notably, the development of TBM has not only allowed the synthesis and purication of difficult peptides but also contributed to the rst successful application of NCL in highly aggregated IL-2 C-terminal part, demonstrating the power of TBM in solving problematic protein synthesis and poor ligation efficiency of insoluble or colloidal peptides. Moreover, the concise installation protocol, in situ on-resin formation, and high compatibility with well-established NCL techniques make TBM an easily adopted approach in difficult protein chemical synthesis.

Results and discussion
As shown in Fig. 2a, IL-2 consists of 133 amino acid residues, and two of the three Cys residues are involved in the disulde bond formation (Cys58 and Cys105). Our initial attempt was to divide the whole sequence into four segments, 1, 2, 3, and 4. The planned synthetic scheme was sequential N-to-C ligation ( Fig. 2b). Firstly, a Ser/Thr ligation (STL) between segments 1 and 2 would provide the ligated peptide thioester 5, then, it would be subjected to the NCL reaction with 3 to afford the product 6. Finally, another NCL would be performed between 6 and 4 to give 7, the desired linear IL-2.
Following this design (Fig. 2b), we rst attempted to synthesize the C-terminal segment 4 via SPPS, but failed  ( Fig. 3a). We noticed that this difficult fragment contained Cys125, a potential site to apply our Cys ligation, as the resultant N,S-benzylidene acetal, the ligation intermediate, has been proven to be powerful for interrupting peptide secondary structure and preventing aggregation. 51 However, the preparation of the corresponding peptide salicylaldehyde ester 4a failed again due to the poor solubility.
Next, we aimed to install a Cys ligation intermediate during Fmoc-SPPS to inhibit the aggregation tendency and enable peptide elongation. To this end, aer the peptide was elongated to Cys125, the sidechain protecting group of the Cys was removed to give a resin-bound peptide 4c (Fig. 3b). Next, a Fmoc-protected Phe salicylaldehyde ester (F-SAL) was dissolved by a pyridine/HOAc buffer (1 : 1) and subjected to react with the resin-bound Cys peptide 4c. Aer overnight reaction, the resin was simply washed by dichloromethane to afford the desired product 4d bearing a Cys ligation intermediate in quantitative yield. Unfortunately, the Cys ligation intermediate was TFA labile, and the N,S-benzylidene acetal was converted to the native Phe-Cys amide linkage during the global deprotection step to afford 4d as its native aggregation form.
To overcome this obstacle, we hypothesized that capping the phenolic hydroxyl group of the Cys ligation intermediate with acetyl group would make it acid-resistant (Fig. 3b) (Fig. 3b). To be specic, the N,S-benzylidene acetal could be deemed to be a tunable switch, i.e., once the phenolic hydroxyl group was masked, it turned into an "off" state to resist TFA deprotection; in contrast, aer Ac removal, it turned into an "on" state and could be removed by TFA.
Moreover, TBM can be easily installed via a very simple onresin in situ generation step, which avoids tedious synthesis. Encouraged by this result, aer the installation of TBM, we performed a standard Fmoc-SPPS and nal Ac capping step to reserve TBM on the sequence to nish the synthesis of 4g.
Indeed, 4 could be successfully synthesized as its TBM form 4g (Fig. 3b), which supported that TBM is a useful strategy to overcome peptide aggregation during Fmoc-SPPS. However, the purication of fragment 4g was unsuccessful because of its high hydrophobicity in aqueous acetonitrile. This preliminary result indicated that disrupting the secondary structure of this IL-2 Cterminal region only prevented its aggregation behavior on the resin and permitted its synthesis, while its hydrophobic nature still hampered the preparation. To alleviate the hydrophobicity, the introduction of a solubilizing tag might be required. Thus, we planned to combine TBM with our reducible solubilizing tags (RSTs) strategy; 50 in other words, the TBM would interrupt the aggregation to guarantee peptide elongation and RST would provide good solubility to enable the purication. With this in mind, Ala112 was mutated to Cys for installation of RST. As shown in Fig. 4a, the synthetic target was revised to 4h.
Naturally, the synthesis was performed by following the same synthetic route of 4g except for replacing Ala112 with Cys. Aer that, the deprotection of the Cys sidechain was carried out under strong reductive conditions to ensure the installation of the solubilizing tag. However, the S-tert-butylthiol protecting group was extremely difficult to remove (Fig. 4a). The reason might be the highly hydrophobic property of this sequence, and the sidechain of Cys was buried by surrounding hydrophobic residues.
As a consequence, we intended to develop an improved version of TBM, which contained both aggregation disruption and solubility improvement capabilities. The hypothesis was to introduce an alkoxy chain at the para position of the aldehyde group of the salicylaldehyde amino acid ester, as shown in Fig. 4b. The alkoxy group would not only render the TBM more acid labile for removal but also provide an opportunity for solubilizing tag installation aer nishing Fmoc-SPPS. Moreover, this design would enable the attachment of the solubilizing tag on the benzene ring of the TBM, which is located at the exterior part of the twisted structure of the peptide mainchain and is less possible to be buried by other residues.
Following the above analysis, SAL ester F-SAL-Alloc was synthesized and subjected to the installation of TBM. Aer the main-chain elongation, the Alloc group was deprotected to allow the attachment of the solubilizing tag (Fig. 4b). Notably, this time, the IL-2 C-terminal region was successfully constructed and puried as 4m in 12% yield, the single, sharp and symmetric peak of this fragment in HPLC suggesting the effectiveness of this TBM strategy in breaking the secondary structure and solving the hydrophobicity issue (Fig. 4c) as hydrophobic peptides normally resulted in broad chromatographic peaks during HPLC and encountered incomplete elution. 68 Moreover, the lack of suitable installation sites of the reported solubilizing tag strategies supports that TBM is a high valuable strategy for the synthesis of the IL-2 C-terminus.
With this success in hand, the other three fragments, 1, 2, and 3, were synthesized in 31%, 55%, and 9% yield, respectively (Fig. 2b). One thing to note was the poor solubility of fragment 3, which required HMB installation at Gly98 for the synthesis and DMSO assistance for HPLC purication. Aer that, STL was performed between 1 and 2 to give the desired ligation product 5, which was subjected to the typical NCL with 3 in a one-pot manner (Fig. 2b). Unfortunately, no reaction occurred aer several optimizations had been tried, including using HPLC puried 5, high reaction concentration, saturated guanidine aqueous solution (8 M), and heating to 37°C. This was attributed to the aggregation tendency of 3 during the reaction.
The failure in the synthesis of 6 prompted us to revise the synthetic scheme. As shown in Fig. 5a, we intended to perform an N-to-C sequential STL between 1, 2a, and 3a to afford 6a, then the Cys ligation between 6a and 4m would provide the desired product 7a, which underwent off-to-on deprotection and acidolysis to give the linear IL-2 7b. Next, fragments 2a and 3a were prepared, and 3a showed similar poor solubility as 3, as expected. With these fragments in hand, the rst STL was performed between 1 and 2a to afford the desired product 5a in 39% yield. Aer that, 5a was ligated with 3a under typical STL conditions to generate the corresponding ligation intermediate, which was later treated with TFA and pyruvic acid to provide the desired SAL ester 6a in 29% yield.
One thing to be noted was the slow formation of the gel in the reaction mixture during the ligation, suggesting the aggregation tendency of 3a in the ligation process. Fortunately, aer ligation, the N-terminal part (1-74) of the IL-2 seemed provide 6a with good solubility to permit further operation. Undoubtedly, 6a was subjected to ligation with 4m according to the synthetic scheme. However, no desired product was observed, even aer changing the reaction solution to pH = 3.0 aqueous buffer containing 8 M guanidine, which normally worked in handling difficult-to-react peptides for Cys ligation.
To address this issue, we hypothesized that although the Nterminal part of 6a provided it with good solubility, the Cterminal still contained the secondary structure, which buried the reactive site and inhibited the ligation. Therefore, to tackle this problematic synthesis and test our hypothesis, we further revised the synthetic scheme (Fig. 5b). The synthesis of the Cterminal region (75-133) was redesigned, and two TBMs were introduced in the construction of fragments 3b and 4n. It was expected that the installation of TBM at Met104-Cys105 would provide 3b with good reactivity and solubility.
Next, 3b and 4n were successfully synthesized by following the protocol established above, and were converted to the corresponding 3c and 4o (Fig. 5b). Remarkably, the TBM strategy exhibited compatibility with the peptide hydrazide chemistry, and permitted hydrophobic and aggregated IL-2 fragments to be synthesized in the hundreds of milligram scale. With these two fragments in hand, an NCL was performed between 3c and 4o to afford the desired product 8. This successful application of TBM in NCL for handling remarkably difficult sequence, particularly the NCL that was performed at a highly steric hindered Thr site, indicated that TBM could totally break the aggregation tendency of such type of difficult peptide and provide good solubility and reactivity.
Aer that, 8 was subjected to a desulfurization protocol to give the corresponding product 8a in 62% yield. This promising result also supported the feasibility of TBM in promoting the solubility and desulfurization of difficult peptides, as it has been proven that poorly solubilized peptides or proteins suffered from sluggish and incomplete desulfurization. 39,48 Finally, the STL between fragments 8a and 5a provided the desired ligation intermediate, which was subjected to TFA treatment for TBM removal and afforded the nal linear IL-2 7 in 21% yield (Fig. 5b). Next, the linear IL-2 sequence was refolded following the reported procedure 54,55 and subjected to HPLC purication to afford folded the IL-2 in 10% yield. The UPLC-MS trace ( Fig. 5c and d) and circular dichroism (CD) spectra (Fig. 5e) supported the folding result of IL-2. 54,55

Conclusion
In conclusion, a TBM strategy has been developed and employed in the synthesis of highly aggregated and hydrophobic C-terminal region of IL-2, and it has led to the successful total synthesis of IL-2. The deprotection or preservation of TBM can be easily adjusted via a simple Ac capping or decapping step, which enables it to fulll the requirement of various kinds of situations during complex protein chemical synthesis. It not only enabled the rst successful application of NCL in the synthesis of such type of difficult peptides but also permitted the desulfurization step. In addition, the concise and efficient installation protocol, in situ on-resin TBM formation, and compatibility with peptide hydrazide preparation approach may promote it as an easily adopted strategy for the chemical synthesis of proteins bearing difficult sequences, such as IL-2. Moreover, the combination of this TBM strategy and thiolderived amino acids, such as penicillamine, may further expand the application scope of TBM to enable a wide range of aggregative and hydrophobic peptides or proteins to be chemically synthesized.

Data availability
Experimental data is available in the accompanying ESI. †

Author contributions
X. L. conceived the idea and supervised the project; H. W. conducted the majority of the experimental work; Y. T. performed the preliminary attempt in the 1 st trial of chemical synthesis of IL-2; W. L. N. helped to prepare some peptide fragments.

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