Visible-light-mediated catalyst-free synthesis of unnatural α-amino acids and peptide macrocycles

The visible light induced, photocatalysts or photoabsorbing EDA complexes mediated cleavage of pyridinium C-N bond were reported in the past years. Here, we report an ionic compound promote homolytic cleavage of pyridinium C-N bond by exploiting the photonic energy from visible light. This finding is successfully applied in deaminative hydroalkylation of a series of alkenes including naturally occurring dehydroalanine, which provides an efficient way to prepare β-alkyl substituted unnatural amino acids under mild and photocatalyst-free conditions. Importantly, by using this protocol, the deaminative cyclization of peptide backbone N-terminals is realized. Furthermore, the use of Et3N or PPh3 as reductants and H2O as hydrogen atom source is a practical advantage. We anticipate that our protocol will be useful in peptide synthesis and modern peptide drug discovery.

P eptides are indispensable bioactive components for exerting biological function normally in various cells. Thus, therapeutic peptide is deemed to be an exquisite substitution of an endogenous molecule which possesses high affinity and selectivity against a pharmacologically diverse set of biological targets 1,2 . Although remarkable achievements have been made in peptide research, poor membrane permeability, low metabolic resistance, and bioavailability still are insurmountable barriers to endogenous bioactive peptide therapeutic 3 . In this context, several methods have been developed and applied to solve these problems, such as incorporation of unnatural amino acids (UAAs) or macrocyclizations of linear peptides 4 .
In recent years, with the rapid development of photochemistry, visible light promoted radical coupling reactions have become important pathway for building chemical bonds 5 . Owing to mild reaction conditions and excellent functional group tolerance, photoinduced chemical transformations provided an excellent strategy for chemoselective biomolecule modification, which are widely applied in the modification of amino acids, peptides, and proteins [6][7][8][9][10][11][12] . The modification of amino acids is an important strategy for preparing UAAs, such as the modification of glycine, cysteine, tyrosine, tryptophan, histidine, etc. Dehydroalanine (Dha) is a naturally occurring amino acid 13,14 , and also can be easily prepared from Ser, Cys, and selenocystein, which is preinserted at the position of interest in peptides and proteins 15,16 . In recent years, Dha has been used as a versatile backbone for synthesis of UAAs. Although these methods provided efficient ways for preparation of various UAAs, the accessing of β-alkyl substituted UAAs were still limited, and the reactions required either transition metal catalysts or stoichiometric amount of metal reagents [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32] . Furthermore, the compatible methods for modification of Dha unit in peptides are still rare 33,34 , and the related peptide cyclization protocol based on functionalization of Dha is still underdeveloped.
The traditional peptide cyclization methods are based on lactamization and disulfide bond formation. In the past decades, with the advent of new generations of peptide pharmaceuticals, varying the nature of ring-forming linkage in peptide macrocycles became a necessity 35 . In this context, transition metal-catalyzed peptide macrocyclization strategies are gaining increasing popularity, and include C-H activation, oxidative cross-couplings, heteroatom ligation, and radical reactions, etc [36][37][38][39] . MacMillan and co-workers recently reported a visible-light-promoted decarboxylative Giese reaction, which provided an efficient way for macrocyclization of peptide backbone C-terminal in an alternative manner besides lactam bond (Fig. 1a) 40,41 .
Primary amines are naturally occurring and chemically diverse starting materials, and deaminative reaction of primary amines has emerged as an important strategy for generating alkyl radicals. Katritzky salts derived from α-1°and α-2°amines, as well as imines derived from sterically encumbered α-3°primary amines are important alkyl radical precusors 42,43 . In line with our interests in photocatalysis and peptide synthesis [44][45][46] , we here report a visible light induced, ionic compounds promoted C-N cleavage of Katritzky salts, which is successfully used in the preparation of β-alkyl substituted UAAs and macrocyclization of peptide. Notably, Et 3 N and PPh 3 are effective to act as singleelectron reductants for the reactions.

Results
Design plan. The Katritzky-type alkyl substituted pyridinium (1) can generate an alkyl radical through a photoinduced dissociative electron transfer 47 52 . Interestingly, in our recent studies, we found that some ionic compounds (M + X -), such as K 2 CO 3 , KOCH 3 , NaI, HCl, CF 3 CO 2 H, CsF, etc., could dramatically increase the absorption of 1 in visible light (400-480 nm) region (Fig. 1b, see the "Supplementary information" for details). The loading amount of ionic compound is directly proportional to the light absorption intensity, indicating the key role of ionic additives for increasing the visible light absorption (The reason that ionic additives increasing visible light absorption of Katritzky salt was unclear at this stage. The mechanistic studies and further applications are underway in our lab). These findings let us wonder whether visible light can be directly used for homolytic alkyl pyridinium (1) C-N bond without assistance of photosensitizers by addition of ionic compounds. If so, upon irradiation, C-N homolysis of 1 can promote the release of alkyl radical. Compared with reported EDA complexes triggered SET to alkyl pyridinium, which were highly dependent on a suitable electron donor (e.g., indoles, Hantzsch esters, or aryl amines) to form a light-absorbing aggregator 44,[49][50][51][52][53] , these ionic compounds promoted photocatalytic C-N homolysis would unveil more rich photochemistry.
To validate our proposal in Fig. 1c, a suitable single-electron reductant (SER) is required. Tertiary amines are previously widely used as SERs. Besides tertiary amines, we speculated that electronrich tertiary phosphines, such as PPh 3 , could also be used in our reaction for electron transfer. We assume phosphine could provide one electron to pyridinium radical cation and form a phosphine radical cation 2, which would then react with H 2 O to give the intermediate 3. The latter could act as an SER again by supplying a second electron and yield phosphine oxide. Overall, the proposed photochemical mechanism to generate C(sp 3 )-centered radicals by using phosphine and H 2 O as SER was not realized before 54 .
Investigation of reaction conditions. For our initial explorations (Table 1), we selected N-Ac-Dha methyl ester (4a) (0.1 mmol) as Michael acceptor and N-Boc-protected cyclic pyridinium salt 1 as alkyl radical precursor. Most of the tested tertiary alkyl amines and trivalent tertiary phosphines could serve as SER for the reaction, while Et 3 N (entry 1) and PPh 3 (entry 2) gave the best results, respectively. Notably, in the case when PPh 3 was used, stoichiometric amount of triphenylphosphine oxide (Ph 3 PO) was formed as a byproduct, indicating that PPh 3 and H 2 O were all involved in SET process. 5a was obtained in 50% yield in the absence of K 2 CO 3 under condition A, which might be attributed to the formation of the EDA complex between Katritzky salt 1 and Et 3 N (entry 3) 52 . It is worth noting that the yield of 5a dramatically increased to 97% in the presence of K 2 CO 3 (entry 1). The yield of 5a decreased to 13% in the absence of KOCH 3 under condition B (entry 4). In addition to K 2 CO 3 and KOCH 3 , fluoride additives could also increase the outcome (entries 5 and 6, respectively). Without Et 3 N or PPh 3 , only a trace amount of desired product formed (entry 7). The reaction gave diminished yield in the absence of H 2 O (entry 8). We assumed that H 2 O might play two roles in the reaction. One is increasing the solubility of an ionic compound; the other is involving in the SET process when PPh 3 is used as an SER reagent (Fig. 1c). Without light irradiation, the reaction did not proceed at room temperature or 60°C(entries 9 and 10, respectively).
Substrate scopes. Having established the optimal reaction conditions (Table 1, entries 1 and 2), we examined a series of alkenes (4b-4l) as shown in Fig. 2a. Similar to Dha, another naturally occurring dehydroamino acid derived alkene, namely dehydrobutyrine, also reacted smoothly to give the corresponding product 5b in 80% yield (4:1 d.r.) under condition A and 76% yield (4:1 d.r.) under condition B. Other alkenes, such as diethyl fumarate, vinyl sulfone, or dimethyl malonate derived alkene were all compatible in the reaction with good yields (5c-5e, 78-95%). The Michael acceptors bearing amide groups also worked well under optimal conditions to give the desired products in uniformly good yields (5f-5j). Furthermore, we expanded the reaction scope to β-thiolated amino acid derivative (4k) which was readily prepared from L-cysteine and dehydroalanine derived Karady-Beckwith alkene (4l). The corresponding hydroalkylation products 5k and 5l were obtained in high yields and good r.r. (rotamers mixture ratio) 55 or d.r. Subsequently, we shifted our attention to the scope of Katritzky salts for the synthesis of unnatural α-amino acids using a series of secondary alkyl substituted pyridinium (Fig. 2b). Uniformly good yields were obtained (7a-7h, up to 92%). Importantly, this reaction could be extended to glycine derived Katritzky salt (7h, 92 and 85%), which provided the possibility for deaminative N-terminal macrocyclization of the linear peptide backbone. Furthermore, more complex pyridinium radical precursors bearing various functional groups and structural motifs were tested under condition A and B, which provided corresponding products 7i-7r in moderate to excellent yields. Unfortunately, unactivated primary alkyl substrates had failed to give the desired hydroalkylation products (When the unactivated primary alkyl substrate was used in the reaction, significant amounts of byproduct from radical-radical coupling of primary alkyl dihydropyridine radical with unactivated primary alkyl radical was obtained. See supplementary information for details.).
Next, a variety of Dha containing peptides was prepared to test the functional group tolerance of this reaction with various amino acid residues incorporated. To our delight, peptides bearing glycine, proline, leucine, tryptophan, phenylalanine, or tyrosine units were able to participate in the deaminative conjugation to furnish the desired adducts in yields between 76 and 95% with  other amino acid residues untouched (Fig. 3a, 9a-9i). Based on excellent performance on modification of peptides, we turned our attention to deaminative N-terminals macrocyclization of linear peptides. As shown in Fig. 3b, a series of peptides that incorporate a structurally diverse set of amino acids can be successfully cyclized using this deaminative method (11a-11e, 28-40% isolated yields after HPLC purifications). Regrettably, the attempt to macrocyclize Dha contained peptide "on resin" during solidphase peptide synthesis procedure has failed.
Mechanistic studies. To gain some detailed information for the reaction, a series of mechanistic studies were carried out (see Supplementary information for details). As shown in Fig. 4a, radical trapping experiments indicated that the combination of ionic compounds (such as K 2 CO 3 , KOCH 3 , and CsF) and pyridinium were able to harness photonic energy from visible light and promoted C-N homolysis of Katritzky salt to generate an alkyl radical. In isotope tracking experiments (Fig. 4b, c), when D 2 O was used instead of H 2 O under Condition A and B, deuterated product 5a-D and 7a-D were obtained in 86 and 34% yield, respectively, revealing that H 2 O acted as hydrogen atom source to provide an H + for the hydroalkylation process. Furthermore, stoichiometric amount of triphenylphosphine oxide (Ph 3 PO) was formed as the byproduct, which suggested that PPh 3 could possibly provide two electrons in a stepwise way during the SET process (Fig. 4b). The UV-Vis absorption of individual reagents or mixtures were shown in Fig. 4d. None of Et 3 N, PPh 3 , pyridinium (1), or the mixture of [PPh 3 + 1] showed significant  , the yield was significantly lower than for K 2 CO 3 promoted reaction (50% vs 97%). Some ionic compounds, such as Cs 2 CO 3 , K 3 PO 4 , and NaOCH 3 , provide higher levels of visible light absorption; however, only moderate yields were obtained (Table S3 and Fig. S8). The use of KBF 4 and NaBF 4 as ionic additives did not give any desired product, which indicated that BF 4 − counter anion in Katritzky salt was not crucial for the transformation (Table S3, Table S13, and Fig. S8).
On the basis of these preliminary results, a possible mechanism is proposed (Fig. 5) (Although the mechanistic investigations suggested that a visible light promoted C-N homolysis of Katritzky salts in the presence of ionic compounds is highly possible in our reaction, the EDA complex pathway cannot be completely ruled out at this stage. See supplementary information for details.). Initially, C-N homolysis of Katritzky salt occurred in the presence of basic ionic compounds under visible light irradiation, followed by the formation of an alkyl radical (•R 1 ) and pyridinium radical cation. Et 3 N (PPh 3 ) was oxidized to nitrogen radical cation A 52 (phosphine radical cation) with the formation of 2,4,6-triphenylpyridine. Notably, under condition A, nitrogen radical cation A is deprotonated to yield α-amino radical, and under condition B, phosphine radical cation reacted with H 2 O to give intermediate B 54 . Radical addition of •R 1 to Michael acceptor 4 gives radical intermediate C. According to the previous report 56 , under condition A, intermediate C was reduced by α-amino radical to give the carbon anion species D and an immonium ion. Subsequently, the hydrolysis of immonium ion promoted the generation of diethylamine and acetaldehyde, followed by the release of a proton. Under condition B, intermediate C was reduced by B to form a carbon anion species D alone with the formation of triphenylphosphine oxide. The final protonation of D provided our desired deaminative hydroalkylation product. In condition B, PPh 3 served as the single-electron reductant and provided two electrons in a stepwise way.

Discussion
In summary, we report ionic compounds promoted homolytic cleavage of pyridinium C-N bond by exploiting the photonic energy from visible light. This finding was successfully applied in deaminative hydroalkylation of a series of alkenes including naturally occurring Dha, which provided an efficient way to prepare β-alkyl substituted UAAs under mild and catalyst-free conditions. Importantly, by using this protocol, the deaminative cyclization of peptide backbone N-terminals was realized. Furthermore, the use of Et 3 N or PPh 3 as reductants and H 2 O as hydrogen atom source is a significantly practical advantage. Further investigating the mechanism of ionic compounds improved visible light absorption of pyridinium and other synthetic applications are currently underway and will be reported in due course.

Methods
General procedure of condition A. To an oven-dried 10 mL quartz test tube with a stirring bar was added alkene (0.2 mmol), alkyl pyridinium salt (0.5 mmol, 2.5 equiv), Et 3 N (0.4 mmol, 2 equiv), and K 2 CO 3 (0.2 mmol, 1 equiv). Then, the air was withdrawn and backfilled with Ar (three times). CH 3 CN (3 mL) and H 2 O (6 mmol, 30 equiv) were added. The mixture was transferred to a violet LED photoreactor (24-W, 410-420 nm), where it was irradiated for 12 h. Then, the reaction was quenched with water (5 mL), extracted with ethyl acetate, washed with brine, dried over anhydrous sodium sulfate, concentrated in vacuo, and purified by column chromatography (hexane/acetone) to afford the product.
General procedure of condition B. To an oven-dried 10 mL quartz test tube with a stirring bar was added alkene (0.2 mmol), alkyl pyridinium salt (0.4 mmol, 2.0 equiv), PPh 3 (0.48 mmol, 2.4 equiv), and KOCH 3 (0.6 mmol, 3.0 equiv). Then, the air was withdrawn and backfilled with Ar (three times). Acetone (3 mL) and H 2 O (5 mmol, 25 equiv) were added. The mixture was transferred to a violet LED photoreactor (24-W, 420-430 nm), where it was irradiated for 12 h. Then, the reaction was quenched with water (5 mL), extracted with ethyl acetate, washed with brine, dried over anhydrous sodium sulfate, concentrated in vacuo, and purified by column chromatography (hexane/acetone) to afford the product.