Commercially available materials were obtained from Adamas, Energy Chemicals, or Sigma-Aldrich. Standard Fmoc-amino acids, 2-chlorotrityl chloride (2-Cl-(Trt)-Cl) resin, 1-hydroxybenzotriazole (HOBt), and 2-(1 h-benzotriazole-1-yl)-1,1,3,3-tetramethyluronium tetrafluoroborate (TBTU) were purchased from GL Biochem (Shanghai). The reagents acetylacetone (acac) was purchased from Aladdin. Ethylene diamine tetraacetic acid (EDTA) and 2,2′-(ethylenedioxy)diethanethiol (DODT) were purchased from TCI. 2,2′-azobis [2-(2-imidazolin-2-yl) propane] dihydrochloride (VA-044) and silver acetate (AgOAc) were obtained from J&K Scientific and innochem, respectively. N,N-diisopropylethylamine (DIPEA), N,N′-diisopropylcarbodiimide (DIC), dithiothreitol (DTT), tris(2-carboxyethyl)phosphine (TCEP), trifluoroacetic acid (TFA), 2-methyl-2-propanethiol (t-BuSH), and guanidine hydrochloride (Gn·HCl) were obtained from Adamas. Methoxylamine (CH₃ONH₂·HCl) and triisopropylsilane (TIPS) were purchased from Energy Chemicals. L-arginine hydrochloride (Arg·HCl), tris(hydroxymethyl)aminomethane (Tris), glutathione reduced (GSH) and glutathione oxidized (GSSG) were purchased from Sangon. The reagents N,N-dimethylformamide (DMF) and dichloromethane (DCM) were purchased from GHTCH (Guangdong). 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU) and 4-mercaptophenylacetic acid (MPAA) were purchased from Alfa Aesar. His₆-Ulp1 was recombinantly expressed using a reported procedure (Malakhov et al., 2004). Acetonitrile (MeCN) used in analytical HPLC and preparative HPLC was obtained from Fisher and Sigma-Aldrich, respectively. Analytical HPLC (Agilent 1,260) was performed on a Phenomenex Jupiter C4 column (4.6 × 250 mm, 300 Å, 5 μm particle size) running at a flow rate of 1 ml/min with UV detection at 214 and 254 nm. Semi-preparative HPLC (Shimadzu AR-20) was performed using a Waters XBridge® peptide BEH C18 OBD™ Prep column, 300 Å, 5 μm, 10 × 250 mm) running at a flow rate of 4.7 ml/min with UV detection at 214 and 254 nm. Preparative HPLC (Ruihe® Tech) was performed using a Welch Ultimate XB-C4 column Prep column, 300 Å, 5 μm, 30 × 250 mm) running at a flow rate of 40 ml/min with UV detection at 214 and 254 nm. Solvent A: 0.1% TFA in water; Solvent B: 0.1% TFA in MeCN. LC-MS was performed on an Agilent LC/MSD (ESI) system on ACE 5 C4 column (150 × 4.6 mm). MALDI-TOF mass spectra (MALDI-8020, Shimadzu) were obtained in the linear positive mode using a matrix of 10 mg/ml α-Cyano-4-hydroxycinnamic acid (HCCA) in water/MeCN (1: 1, v/v) with 0.1% TFA.

SelF(1–41)-MPAA (1) was synthesized on 2-Cl-(Trt)-Cl resin (theoretical loading: 0.9 mmol/g) using Fmoc-Gly-OH with 0.4 mmol/g loading and elongated according to standard Fmoc-SPPS protocols highlighted in Preloading of 2-Cl-(Trt)-NHNH₂ Resin, Estimation of Amino Acid Loading, Fmoc Deprotection, Coupling of General Amino Acids, and Coupling of Fmoc-Cys(Trt)-OH or Fmoc-Cys(Acm)-OH to afford resin-bound peptide. The peptide was cleaved using TFA/H₂O/DODT (95:2.5:2.5 v/v/v) for 2 h, and according to Cleavage of the Crude Peptide to acquire the crude peptide SelF(1–41)-NHNH₂ (1a). The crude peptide 1a (400 mg, assumed 100% purity) was dissolved to 40 mg/ml in 6 M Gn·HCl, 0.2 M Na₂HPO₄, pH 3.0, with 10 equiv MPAA, 2.5 equiv acac (from a 0.1 M stock in water) were added to the mixture, and the reaction mixture was stirred for 10 h to form thioester fragment SelF(1–41)-MPAA (1). The mixture was centrifuged (8,000 rpm, 5 min, 4°C), filtered and purified by preparative HPLC at 25°C with a gradient of 25–70% MeCN (with 0.1% TFA) in 25 min to obtain 8 mg of segment 1 (1 g resin; 2.5%). Overall, 18 mg of segment 1 was obtained. The purity and exact mass of the peptide was confirmed using analytical HPLC and ESI-MS, respectively (Supplementary Figure 1).

SelF(42–74)-MPAA (2) was synthesized on 2-Cl-(Trt)-Cl resin (theoretical loading: 0.9 mmol/g) using Fmoc-Ala-OH with 0.35 mmol/g loading. Through a similar procedure described in Synthesis of Thioester Fragment 1 and a preparative HPLC at 25°C with a gradient of 25–50% MeCN (with 0.1% TFA) in 25 min, 25 mg of segment 2 was obtained (1 g resin; 7.4%). Overall, 75 mg of segment 2 was obtained. The purity and exact mass of the peptide was confirmed using analytical HPLC and ESI-MS, respectively (Supplementary Figure 2).

The gene encoding for the His₆-SUMO-SelF(75–134) fusion protein was synthesized and codon-optimized for E. Coli expression (GenScript Inc., Nanjing). The synthetic gene was cloned into the pET-30a expression vector using the NdeI/EcoRI restriction sites. The plasmid was firstly transformed into BL21 (DE3) E. coli cells chemically. An overnight culture of the cells harboring an expression vector was inoculated (1:50 dilution) in a 4 L flask containing 30 µg/ml Kanamycin in 2 L LB at 37°C. After reaching an OD₆₀₀ of 0.6–0.8 overexpression of His₆-SUMO-SelF (75–134) was induced by the addition of 2 ml 1 M IPTG stock solution (final conc. 1 mM) at 37°C for 4 h. Cells were harvested by centrifugation (8,000 rpm, 4°C, 15 min). Typically, 6 g cells were resuspended in 50 ml of cell lysis buffer and lysed by ultrasonication (30–40% power, 3 s on 5 s off, 25 min). The crude lysate was centrifuged (16,000 rpm, 4°C, 20 min) and the supernatant was discarded. The precipitate was stirred at 4°C overnight with 10 ml of Ni-NTA binding buffer to extract His₆-SUMO-SelF(75–134). The precipitate was removed by centrifugation (16,000 rpm, 30 min, 4°C, 5 cycles) and the supernatant applied to a HisTrap™ FF column (5 ml) at 2 ml/min with a AKTA pure chromatography system. Absorption was monitored at 280 nm. The column was washed with 20 ml of Ni-NTA binding buffer (see Supplementary Material for details). His₆-SUMO-SelF(75–134) was eluted with 25 ml of Ni-NTA eluting buffer (see Supplementary Material for details) in fractions of 5 ml. The 10 ml fractions with A280 > 0.1 (Supplementary Figure 3A) was gradually dialyzed to the 0.5 L refolding buffer (see Supplementary Material for details) to complete the refolding of SUMO domain. 400 µL of a stock solution of His₆-Ulp (A280 = 0.5) were added to the folded His₆-SUMO-SelF(75–134) (3a) (10 ml, A280 = 1.8) (Ulp1: protein = 2:50, v/v) and the reaction was incubated for 2 h at 30°C. Equal volume of the buffer containing 6 M Gn·HCl, 200 mM Na₂HPO₄, 200 mM CH₃ONH₂·HCl, 10 mM TCEP (pH 3), was added to the mixture and the pH was adjusted to 4.0 to remove any C-terminal cyclized byproduct. After overnight incubation, the mixture was centrifuged (8,000 rpm, 5 min, 4°C), filtered and purified by preparative HPLC at 25°C with a gradient of 25–48% MeCN (with 0.1% TFA) in 20 min to obtain 13 mg of SelF(75–134) (3) (6.5 mg/L LB) (Supplementary Figure 3C). The purity and exact mass of the peptide was confirmed using analytical HPLC and ESI-MS, respectively (Supplementary Figures 3C-c,D).

The peptide-thioester 1 (0.76 μmol, 2 equiv, 3.48 mg) and Cys-peptide 5 (0.38 μmol, 1 equiv, 4 mg) was dissolved in 190 μL of the ligation buffer (6 M Gn·HCl, 200 mM Na₂HPO₄, 20 mM TCEP, 50 mM MPAA, pH 6.5). The solution was incubated at RT for 24 h (confirmed by LC-MS monitoring). After the conversion was completed, the ligation reaction mixture was reduced by the addition of 2.86 mg TCEP (final conc. 50 mM) and incubated for 15 min. Then, 200 μL of the ligation mixture (in 50 μL portions) was exchanged into a 150 μL buffer containing 6 M Gn·HCl and 200 mM Na₂HPO₄ at pH 6 (Amicon® Ultra-0.5 concentrator, 0.5 ml, 3 K MWCO, 11,000 rpm, 4°C). The resulting ligation mixture (∼600 μL) was added dropwise to a 20 ml refolding buffer (0.4 M Arg·HCl, 0.2 M Tris, 0.1 M (NH₄)₂SO₄, 2 mM EDTA, 0.2 mM GSSG, 1 mM GSH, pH 8.2) (final protein conc. was ∼0.25 mg/ml) at 4°C in 30 min, and was stirred in a refrigerator at 4°C for 12 h. The mixture was centrifuged (8,000 rpm, 5 min, 4°C), filtered and purified by semi-preparative HPLC with a two-step gradient: 10–30% in 5 min, then 30–55% MeCN (with 0.1% TFA) in 30 min to collect the desired fractions and lyophilized, and the desired folded protein SelF(U65C/Q74A) (7) as a white amorphous powder (1 mg, 17.6% isolated yield over two steps). The purity and exact mass of the peptide was confirmed using analytical HPLC and ESI-MS, respectively (Supplementary Figures 6A-e,C).

The gene coding for the His₆-SUMO-SelF(U65C) fusion protein was synthesized and codon-optimized for E. Coli expression (GenScript Inc., Nanjing). The synthetic gene was cloned into the pET-30a expression vector using the NdeI/EcoRI restriction sites. Site-directed mutagenesis using the Sit-directed Mutagenesis kit from Sangon (Shanghai) was carried out to afford the two Cys-to-Ser variants, U65C/C42S and U65C/C43S respectively. The resulting expression plasmids were transformed into E. coli BL21 (DE3). In a typical protein expression experiment, cells were grown in a 2-L LB medium containing 30 ug/mL of kanamycin at 18°C. Expression was induced after reaching an OD₆₀₀ of 0.6 with 1 M IPTG stock solution (final conc. 0.5 mM) and cells were grown for 20 h at 18°C. The cells were harvested by centrifugation (8,000 rpm, 4°C, 15 min). Typically, 4.5 g of cells were resuspended in 30 ml of cell lysis buffer and lysed by ultrasonication (30–40% power, 3 s on 5 s off, 25 min). The crude lysate was centrifuged (16,000 rpm, 4°C, 30 min, 5 cycles) and the supernatant applied to a HisTrap™ FF column (5 ml) at 2 ml/min with a AKTA pure chromatography system. Absorption was monitored at 280 nm. The column was washed with 20 ml of Ni-NTA binding buffer (see supplementary material for details). The desired proteins were eluted with 25 ml of Ni-NTA eluting buffer (see supplementary material for details) in fractions of 5 ml. The 10 ml fractions with A280 > 0.1 (Supplementary Figures 7, 8, 9A) were collected. Then, 400 µL of a stock solution of His₆-Ulp (A280 = 0.5) was added to the fusion protein (8a, 9a or 10a) (10 ml, A280 = 2.1) (Ulp1/protein = 2/50, v/v) and the reaction was incubated for 2 h at 30°C. An extra alkylation step was carried out for the two variants by adding 100 mg of iodoacetamide (IAM) (10 mg/ml) and incubating for 1 h to protect the free Cys residue with a S-carboxyamidomethyl (CAM) group. The mixture was centrifuged (8,000 rpm, 5 min, 4°C), filtered and purified by preparative HPLC using at 25°C with a gradient of 30–55% MeCN (with 0.1% TFA) in 30 min to obtain the desired proteins: 8–9 mg (4–5 mg/L LB) (Supplementary Figures 7, 8, 9C). The purity and exact mass of the protein was confirmed using analytical HPLC and ESI-MS, respectively (Supplementary Figure 7C–D, 8C–D, 9C–D).

The secondary structure content of the synthetic SelF(U65C/Q74A) (7) was compared to the recombinant SelF(U65C) (8) using far-UV CD spectroscopy (200–260 nm). Spectra were recorded on a Chirascan™-Plus Circular Dichroism Spectrometer (Applied Photophysics Ltd., United Kingdom), using a quartz cuvette with a path length of 0.1 cm, and obtained by averaging 3 wavelength scans in 1 nm steps, with a signal averaging time of 1 s and a bandwidth of 1 nm. Each purified protein was dissolved separately in NH₄HCO₃ buffer. Measuring conditions: Protein conc.: ∼38 µM; Buffer: 10 mM NH₄HCO₃, pH 8. The folding of the expressed proteins was also confirmed by CD spectroscopy, which gave very similar spectrum for each protein (Supplementary Figure 10).

Protein was dissolved in a solution of 25 mM NH₄HCO₃ (1 mg/ml, 100 µL). Trypsin (0.5 mg/ml, 10 µL) was firstly added to the mixture to digest the protein (Diemer et al., 2020). The reaction was carried out at 37°C for 2 h. Then, the enzymatic reaction was analyzed by LC-MS directly. Next, chymotrypsin (0.16 mg/ml, 11 µL) was added to the reaction mixture for further digestion. The reaction was carried out at 30°C for 2 h. Finally, the enzymatic reaction was analyzed by LC-MS.

A major aim of the current work is to establish an efficient synthetic route for the Cys homolog of the mature human SelF protein—which could later be applied for the synthesis of the Sec-containing protein. For this purpose, we initially used a three peptide segments semi-synthetic strategy, in which a short peptide hydrazide segment Cys65–Gln74 was synthesized by SPPS. The remaining two segments could be obtained from recombinant expression. This way we hoped to maximize the synthetic yield of the peptide segments and thus the synthetic protein. An extra C-terminal Cys residue was included in the expression plasmid for the first segment—Phe–Gly64, which was supposed to be able to transform to the corresponding thioester via a reported hydrazinolysis procedure (Adams et al., 2013; Pan et al., 2019). However, the effect of hydrazinolysis is not satisfactory, giving only minor product (data not shown). We thus switched to an alternative strategy where disconnection sites at Gly41–Cys42 and Ala74–Ala75 were adopted. It is noted that for the purpose of applying the peptide hydrazide chemistry, the C-terminal Gln74 residue in the middle segment was mutated to an Ala residue, which should not affect the protein folding and function (vide infra). Here, segments 1 and 2 can be obtained directly from SPPS, while segment 3 can be fused N-terminally to a His₆-SUMO tag and recombinantly expressed. As shown in Figure 1, the peptide segments were designed to be ligated by native chemical ligation (NCL) (Dawson et al., 1994), and thus an N-terminal Cys residue mutation will be installed in 3 and after its ligation with segment 2, a desulfurization could lead to the corresponding Ala residue (Wan and Danishefsky, 2007). For this purpose, all Cys sidechains of segment 2 have to be protected with Acm in prior, which after the ligation and desulfurization can be removed. The resulting segment can then be ligated with segment 1 to afford the desired full-length protein 6.

As mentioned, peptide segments 1 and 2 were obtained via SPPS, and C-terminal α-hydrazide was used as thioester precursor. The use of pseudoproline dipeptide Fmoc-Phe-Ser(Ψ^Me,Mepro)-OH (Wöhr, et al., 1996) proved to be essential for a reasonable yield (Figure 1A). Both segments were cleaved from the resin as peptide hydazides (Fang et al., 2011) and the crude products were transformed into peptide-MPAA thioesters via an acetyl acetone (acac) method before HPLC purification (Figures 2A,B) (Flood et al., 2018). It is noted that NCL with purified peptide thioesters gave better yields than in situ method applying peptide hydrides (vide infra) Segment 3 was obtained recombinantly with an N-terminal His₆-SUMO tag (Figure 2C) (Malakhov et al., 2004; Reif et al., 2014). And after Ulp1-cleavage, the desired segment 3 was obtained with high purity and yield (6.5 mg/L LB, Figure 2D). It is noted that before HPLC purification, the Ulp1-cleavage product was treated with methoxylamine to remove any cyclized thiazolidine byproduct at the N-terminus (Reif et al., 2014).

With all the peptide in hand, we preceded to assemble the full-length protein via NCL. The first NCL was carried between peptide segments 2 and 3, and with the purpose of applying one-pot desulfurization where the use of MPAA as catalyst is problematic, imidazole was instead used as the catalyst for the ligation (Sakamoto et al., 2016). As shown in Figure 3A,B, the ligation went almost to completion within 1 h, and the reaction progress is better overviewed from the SDS-page analysis (Figure 3C). To this mixture, TCEP and VA-044 were added directly to achieve quantitative desulfurization at Cys75, leading to peptide 4 with excellent isolated yield (29.1 mg, 53.8% over two steps). And after HPLC purification, the Acm protection groups can be conveniently removed using AgOAc (Murar et al., 2020), affording 5 with high isolated yield (14.2 mg, 49.0%, Figure 3A, d).

Further, a second ligation between peptide 1 and 5 was carried out to obtain the full-length synthetic protein. In this case, MPAA was added as the catalyst and the ligation proceeded smoothly, albeit with a slow reaction rate compared to the first ligation. The presence of thiolactones (indicated with * in Figure 4A, a-b) resulting from the thiol-exchange between the C-thioester and thiol side-chains in peptide 1 was the key for the slow reaction rate. The ligation was allowed to proceed for 24 h and the resulting solution was directly exchanged into refolding buffer, according to the literature (Vetter et al., 2009) (Figure 4A-C). Gratifyingly, a sharp peak with high intensity in HPLC appeared after 12 h, which as judged from ESI-MS analysis (∼–8 Da compared to the unfolded protein, 6) corresponds to the folded synthetic protein 7. Following semi-preparative HPLC purification, 7 was obtained with an isolated yield of 17.6% (3 mg, over two steps, ligation and refolding). Finally, the folded synthetic protein 7 shows identical CD spectrum to the expressed protein counterpart (Figures 4D,E), confirming the correct formation of the 3D structure (Dery et al., 2017).

As mentioned earlier, the disulfide pairing mode of SelF has not been established. The elucidation of such a pairing mode would shed light on its overall structure, especially when the 3D structure of the UGGT binding domain is currently not known. Herein, we identified all disulfide bonds in the SelF(U65C) through a two-step enzymatic digestion of both synthetic and expressed proteins. As shown in Table 1, following trypsin treatment of the recombinant protein SelF(U65C), 8, a peptide fragment bearing the Cys63–Cys65 disulfide can be clearly observed (Entry 1) (Supplementary Figure 12A), which agrees well with the reported NMR structure of a homolog protein—the fruit fly Sep15 (Ferguson et al., 2006). Further treatment of the peptide mixture with chymotrypsin allows the identification of a peptide fragment containing Cys21–Cys24 (Table 1, Entry 3) (Supplementary Figure 12B). The connectivity mode of the remaining four Cys residues is, however, difficult to resolve as Cys42 and Cys43 are next to each other. In this case, we created two Cys-to-Ser variants, i.e. SelF(U65C/C42S) and SelF(U65C/C43S), and the now-free Cys in each variant was firstly protected with IAM. The resulting proteins—SelF(U65C/C42S)-CAM, 9 and SelF(U65C/C43S)-CAM, 10—were then processed through a similar sequential trypsin/chymotrypsin digestion, and the remaining two disulfide bonds were established as Cys10–Cys43 and Cys39–Cys42, respectively (Table 1, Entries 4 and 7) (Supplementary Figure 13A, 14B). Importantly, the synthetic protein showed the same disulfide pairing mode as the expressed protein (Supplementary Figure 11, 12), which reassures the viability of the synthetic strategy developed in the current work to provide authentic samples for further biological studies.