Tyrosine and Drug-like Late-stage Benzylic Functionalization via Photoredox Catalysis

Visible light mediated late-stage functionalization is a rising eld in synthetic and medicinal chemistry, allowing the fast and diversied modication of valuable, potentially therapeutic compounds such as peptides. However, there are relatively few mild methodologies for the C(sp 3 )-H functionalization of complex peptides. Herein, we report a visible light mediated photocatalytic protocol for the benzylic C-H modication of tyrosine and related C-H bonds. The embraced radical-cation/deprotonation strategy enables an incorporation of a wide range of valuable functional groups in high yields and chemoselectivity. The mild reaction conditions, site-selectivity and high functional group tolerance was highlighted by the functionalization of complex peptides, drugs and natural products, providing a promising synthetic platform in medicinal chemistry.


Introduction
Late-stage functionalization (LSF) of C-H bonds, which enables the incorporation of functional groups at the end of a synthesis sequence or in natural products, is an emerging eld in organic synthesis and drug discovery. [1][2][3] The rapid access to structurally diverse chemical series has a signi cant impact on the time and cost consuming optimization of clinical candidates in drug discovery. 3 Nevertheless, due to the ubiquity of C-H bonds, the chemo-and site-selectivity are still major challenges in LSF. 4 Recently, LSF via photoredox catalysis has gained great attention. [5][6][7][8] In this regard, various selective photocatalyzed LSF methodologies, e.g. for the functionalization of C(sp 2 )-H bonds, [9][10] α-heteroatom positions, [11][12][13] or decarboxylation reactions, [14][15] have been developed. However, benzylic photocatalyzed LSF is signi cantly underrepresented. 5 In particular, the benzylic functionalization of electron rich compounds such as in tyrosine or other phenolic drug-like derivatives is highly appealing, since this type benzylic C-H bonds are ubiquitous in biomolecules and pharmaceuticals ( Fig. 1, a). 16 A commonly used approach in photoredox C-H functionalization relies on hydrogen atom transfer (HAT), [17][18][19][20][21][22][23][24][25] in which a photocatalytically generated radical species is able to abstract a hydrogen atom from the substrate. However, C-H bond discrimination issues often lead to a poor site-selectivity for benzylic positions. 5 In fact, other C-H bonds like in tertiary or α-heteroatom positions are modi ed under similar HAT conditions, which compromises the potential for benzylic LSF. [23][24][25][26] Therefore, alternative methods to overcome the current reactivity and selectivity issues are highly demanded. On course of our research program on amino acid and peptide modi cation, 13,27−28 we imagined that a photocatalytic radical-cation/deprotonation strategy [29][30][31][32][33] could be employed as a more promising technique for the selective functionalization of tyrosine derivatives, as well as related phenolic drugs ( Fig. 1, b). In this approach, a reactive radical cation is formed upon single electron transfer (SET) to the excited photocatalyst. Due to a signi cant increased acidity of the benzylic position in the radical cationic species (pk A ~ − 13), 34 a deprotonation to form a benzylic radical is possible, which can then be further modi ed by reaction with a suitable acceptor. The selectivity of the method strongly depends on the redox potentials of the photocatalyst and the substrate, and thus a predictable site activation can be achieved.
Although, this strategy has been e ciently employed in benzylic C(sp 3 )-H modi cations of simple substrates, [29][30][31][32][33] it has been long neglected for LSF. [35][36][37] Thus, a late-stage C-C coupling of electron-rich benzylic positions is unknown embracing this promising methodology. Taking all into account, the development of a novel site-selective LSF methodology of phenol-type structures is highly desirable as it would represent a powerful tool in synthetic and medicinal chemistry. Hence, we herein present a mild photocatalytic functionalization methodology of benzylic C-H bonds in tyrosine derivatives, which can be extended for LSF to other related C-H bonds in peptides and tyrosine-like phenolic drugs and natural products Results

Reaction Optimization
In this regard, we started our studies with the examination of the reaction of N-acetyl tyrosine methyl ester (1, E p/2 = + 1.58 V vs SCE in CH 3 CN, see S.I.) as the model substrate with methyl acrylate (2a) as the radical acceptor in acetonitrile at room temperature upon irradiation with blue LEDs in a photoreactor (Table 1; see S.I. for full screening). As expected, the control experiments without catalyst and/or light were negative, illustrating the need of a photoredox catalyst (entry 1). Therefore, a preliminary screening of photocatalysts was made, showing that photosensitizers with a redox potential slightly above of the one of tyrosine were more effective. In particular, the iridium-based photocatalysts with the dCF 3 bpy ligand (E* p/2 = +1.65 V 38 and +1.68 V 38 vs SCE for PC5 and PC6, respectively) led to the best results (entries 6 and 7), while diminished conversions were obtained with the acridinium PC2 (E* p/2 = +1.65 V vs SCE) 39 (entry 3). Photocatalysts with signi cant higher potentials like the Fukuzumi catalyst (PC1, +2.06 V vs SCE; 39 entry 2) or with potentials below of the tyrosine like 4-CzIPN (PC3, E* p/2 = +1.43 V vs SCE; 40 entry 4) or Ir-4,4´-dtbpy (PC4, E* p/2 = +1.21 V vs SCE; 41 entry 5) showed no conversion. Systematic reaction screening led to the identi cation of [Ir(dFCF 3 ppy) 2 (4,4´-dCF 3 bpy)]PF 6 (PC5) (2 mol%) in CH 2 Cl 2 (0.2 M) as the optimal system (entry 9). However, side product analysis revealed the dimerization of the tyrosine and oligomerization of methyl acrylate as major side-reactions. The observed oligomerization side reactions indicated an ineffective closing of catalytic cycle (E p/2 = -0.79 V vs SCE). 38 Therefore, to hamper the dimerization reaction, the radical acceptor loading was increased to 5.0 equivalents (entry 10). Although only a slightly improvement in the yield (from 41 to 48%) was achieved, the formation of the dimerization product was drastically suppressed. blue light irradiation at 20 °C), different radical acceptors were next explored with 1 as substrate (Fig. 2, a). Aiming at further improving the conversion of this process, the more reactive radical acceptor 1,1bis(phenylsulfonyl)ethylene (2b) was rst explored. Pleasantly, the desired product was afforded in an excellent yield of 96% and a good diastereomeric ratio (d.r. 3:1). Moreover, no oligomerization was observed and only low amounts of difunctionalized product was formed after elongated reaction time.
The con guration of the major diastereoisomer of product 3b was determined as (2S,3S) by NOESY experiments (see S.I.). Similarly, other activated alkenes like benzylidene malonitrile (2c) or electron acceptors with aryl sulfonyl leaving groups 2d-f [42][43][44] could be e ciently enrolled in the reaction. As a result, valuable cyano, allyl and alkynyl groups could be selectively introduced in excellent yields (up to 96%) and good diastereoselectivities (up to 6:1 d.r).
Considering the importance of orthogonal protecting group strategies in peptide chemistry, 45 the in uence of different protecting groups often used in medicinal chemistry and drug-target discovery was studied next (Fig. 2, b). Alkylic O-protecting groups, such as methyl, tert-butyl or benzyl afforded the benzylic substitution products in excellent yields (3-5, up to 96%). In case of the O-silyl protection (TBS), a synthetically useful yield of 69% (6b), was observed. N-acyl protecting groups like acetyl or carbamates such as Boc, Fmoc or Cbz were well tolerated, leading to the desired products 7-9 in good yields. Additional to methyl esters, tert-butyl and benzyl esters were also accepted, providing the products in excellent yields (10b and 11b, 87 and 91%, respectively).
The synthetic utility of the methodology was subsequently illustrated with synthesis of novel tyrosine amino acids by derivatization of the bissulfonyl product 3b (Fig. 2, c). Tyrosine 12 was obtained in an excellent 93% yield after deprotection of the methyl ether upon addition of TMSI. Moreover, the reductive desulfonylation by the Mg/MeOH system afforded the ethyl tyrosine analogue 15 in 84% yield and with no change in the diastereomeric ratio. By uorination of 3b with select uor prior to the reduction, valuable mono-uorinated ethyl tyrosine 16 was obtained in a good yield. The bisphenylsulfonyl group could be also converted to an ester moiety by oxidation with mCBPA, which cyclizes to the lactame 14.
Functionalization of tyrosine-containing peptides Next, the methodology was investigated for the functionalization of complex tyrosine-containing peptides (Fig. 3). In this regard, a selective tyrosine functionalization in benzylic position was obtained in good to excellent yields in peptides with up to 8 amino acid units. Other possible reactive sites like tertiary C-H bonds (Val, Ile, Leu), other benzylic positions (Phe), α-to heteroatom positions (Pro, Lys, Thr) or αC-H glycine bonds remained untouched under the applied conditions, which provides a complementary strategy to previous reported HAT functionalization approaches. 13,23,26 Noteworthy, the con guration analysis on dipeptides revealed an opposite con guration at the newly formed stereocenter compared to the single amino acid 3b. Hence, the major isomer of the dipeptide 17b was determined as the (2S, 3R) (see S.I.). Moreover, other alkyl and alkynyl functional groups could also be introduced in moderate to very good yields.

Late-stage benzylic functionalization of drugs
Considering that tyrosine is an important metabolic precursor, [46][47][48] especially in the synthesis of neurotransmitters (dopamine, adrenaline, etc.) or other secondary metabolites, we were interested in applying our methodology directly to this type of substrates (Fig. 4). To our delight, dopamine and the Parkinson drug L-Dopa were converted to the corresponding products 22 and 23 in excellent yields. Moreover, due to a higher observed reactivity, the amount of radical acceptor could be decreased to 3.0 equivalents.
Additionally, unnatural α-aryl substituted derivatives can also participate in this reaction upon elongated reaction times (≥72 h). Consequently, α-quaternary carbon containing compounds such as 24b could be obtained in good yield. Furthermore, natural products like the opium alkaloid papaverine or the secondary metabolite catechin were successfully modi ed under these conditions, providing the alkylated products 25c and 26b in good to very good yields. Interestingly, the catechin analogue was obtained as a single diastereomer with exclusive functionalization of the secondary benzylic center. Aiming at a broader applicability, we further focused our attention to additional phenol moieties, which are encountered in many dugs. In this regard, blockbuster drugs, like gem brozil or indomethacin, and the insecticide etofenprox were transformed to the corresponding products 27-29 in good to excellent yields. In case of the functionalization of primary benzylic positions (gem brozil and indomethacin), the amount of radical acceptor was decreased to 1.1 or 1.5 equivalents in order to prevent dialkylation. Moreover, a selectivity control was also observed from the used radical acceptor. Consequently, the reaction of the sterically hindered benzyliden malonitrile (2c) with indomethacin showed exclusive mono-functionalization of the primary, electronically richer position (product 28c). A similar observation was made in the modi cation of gem brozil presenting two primary benzylic positions. A favored functionalization of the more activated ortho over the meta methyl group was observed with 2c (product 27c). However, a small amount of the less favored meta-benzylic alkylation was observed with the more reactive bissulfonyl alkene 2b (27b, r.r. >18:1).

Mechanistic investigation
Finally, aiming at shedding some light into the photoredox catalytic process of this reaction, mechanistic investigations were carried out. Stern-Volmer quenching experiments showed a strong interaction of the excited photocatalyst with the tyrosine derivative 1 and no quenching with the alkene 2b (Fig. 5). Interestingly, the product 3b was found to be a good quencher too, though no difunctionalized products could be detected. Presumably, the possible di-functionalization is suppressed due to steric effects.
Additionally, a low quantum yield (F) of 0.03 was measured (see S.I.), which might most likely exclude a radical chain reaction.
Taking all the previous observations into account, a proposed mechanism for the photocatalytic oxidative C-H functionalization of tyrosine with alkenes as radical acceptors is outlined in Fig. 6. Upon light excitation of the photoredox catalyst, tyrosine is oxidized to the corresponding radical cation II by single electron transfer to the excited photocatalyst [Ir*], following a reductive quenching cycle. Next, deprotonation of II to form the benzylic radical III, followed by radical addition to the alkene lead to the radical intermediate IV. The catalytic cycle is closed by electron transfer from the reduced photocatalyst species [Ir •-] to the radical IV, forming the anionic species V. Then, the nal product VI is obtained after protonation. In order to determine the source of protons for this last step, the reaction was performed in deuterated dichloromethane. However, no deuterium incorporation in the product was observed, indicating the tyrosine substrate as the proton source.

Discussion
In summary, we have developed a novel visible light-mediated functionalization strategy of electron-rich benzylic C-H bonds in tyrosine containing peptides and drug-like compounds. In this photoredox catalyzed radical-cation/deprotonation approach, a broad range of valuable functional groups could be selectively introduced into the tyrosine backbone. The combination of mild reaction conditions, chemoselectivity and functional group tolerance was highlighted by the late-stage functionalization of complex peptides, drugs and natural products. Hence, the demonstrated mild and selective generation of benzyl radicals in complex molecule scaffolds provides a powerful strategy with broad applicability in synthetic and medicinal chemistry.

Data Availability
The authors declare that the data supporting the ndings of this study are available within the article and Supplementary Information le, and also are available from the corresponding author upon reasonable request.

Declarations
Acknowledgements (optional) INTERREG V A, ETZ 2014-2020 -Bayern-Czech Republic (project 41) is gratefully acknowledged for generous support. We want to thank Remo Weck for experimental support.
Ethics declarations