Effective bioconjugation reagents are chemoselective for specific residues, amenable to mild and aqueous reaction conditions, and able to perform reactions at low concentrations, which necessitates favorable reaction kinetics. Cysteine presents itself as a common target for site-selective bioconjugation due to its high nucleophilicity, relatively low abundance in the proteome, and ability to be incorporated into proteins, antibodies, and other biomolecules through molecular biology techniques. [1][2][3] Commonly used cysteine modifications rely on addition to maleimides or other Michael acceptors. However, these linkages are susceptible to retro-Michael reactions under reducing conditions, leading to reversible modification of the biomolecule. [4][5][6][7][8][9] Accordingly, cysteine S-arylation has emerged as a unique approach to produce C(sp 2 ) -S bonds, which are less susceptible to reversibility. 10 Many cysteine S-arylation approaches rely on nucleophilic aromatic substitution (SNAr), which necessitates electron withdrawing groups ortho or para to a leaving group. [11][12][13][14][15][16][17][18][19] This limits the aryl substrate scope for potential modification. An alternative method uses metal-mediated or metalcatalyzed cysteine S-arylation, which allows for the ex situ or in situ formation of an oxidative addition complex (OAC). [20][21][22][23][24][25][26][27][28][29][30] These OACs are then able to permit initial thiol coordination to the metal center and subsequent reductive elimination to form the stable, C(sp 2 )-S bond. The use of transition metals allows for the aryl scope to be broadly expanded to contain both electron donating and
Please do not adjust margins Please do not adjust margins withdrawing substituents as well as different steric substitution patterns.
The ex situ formation of the OAC can provide a set of bench stable reagents that can be used for subsequent biomolecule transformations without any additional precautions to preserve the metal reagent. 31,32 By using thiophilic metal centers (e.g. Pd and Au), exquisite chemoselectivity for thiols and rapid kinetics are observed. Notably, the mechanism for the S-arylation proceeds via a two-step process of an initial coordination of the thiol to the metal center followed by reductive elimination to form the S-aryl bond. This stands in contrast to other cysteine modification mechanisms that rely on displacement of an alkyl halide or 1,4-addition into a Michael acceptor. Our groups have reported the use of P,N-ligated Au(III) OACs for the S-arylation of cysteine (Fig. 1), [33][34][35] and recently, we have shown that stopped-flow/UV-vis spectroscopy is a powerful technique to characterize the rates of the two distinct elementary steps of the S-arylation. This allowed for the quantification of the bimolecular thiol coordination rate constant (kCo) for P,N-ligated Au(III) OACs with di-1-adamantylphosphine (PAd2) 1 and di-tertbutylphosphine (P t Bu2) 2 P-substituents as 2,460 and 4,220 M -1 s -1 , respectively. The clear steric dependence on this coordination event led to the synthesis of a dicyclohexylphosphine-based (PCy2) P,Nligated Au(III) OAC reagent 3 with an unprecedented kCo of 16,600 M -1 s -1 . 35 By increasing the steric bulk of the aryl group, the coordination of the thiol to the Au(III) center was slowed down, which allowed for the chemoselective S-arylation of cysteine. By calculating the percent buried volume (%VBur) of the Au(III) OAC, a steric parameter was produced that could be used to predict the rate of coordination to the Au(III) center for both the ligand and aryl modifications. Herein, we demonstrate the extrapolation of %VBur as a screening metric for the expansion of Au(III) OACs for S-arylation. As the dialkylphosphine and amine could both be modified during the synthesis of the P,Nligand, in silico screening was performed for eight dialkylphosphine substituents and five amine groups to examine their steric properties (Fig. 1, 4). Additionally, the free energy of reaction (∆GRxn) as well as the free energy of activation for the reductive elimination (∆GRE ‡ ) were calculated to examine the thermodynamics of each P,N-ligated Au(III) OAC. This allowed for the rapid screening of 13 different Au(III) OACs. Three of these novel P,N-ligated Au(III) OACs were synthesized, then their kinetics and efficiency as S-arylation reagents were determined. This led to the generation of kCo up to 20,200 Please do not adjust margins Please do not adjust margins M -1 s -1 , improving upon the previously developed Au(III) OACs and demonstrating the computational and experimental synergy available in the investigation of metal-mediated bioconjugation reactions. These results also provide additional mechanistic insight into the coordination chemistry of Au(III) compounds. [36][37][38][39][40]
Computational Screening of P,N-Ligated Au(III) OACs Previously, DFT calculations for the S-arylation of methanethiol were performed with NMe2-containing P,N-ligated Au(III) phenyl OACs with PAd2 (5), P t Bu2 (6), and PCy2 (7) P-substituents. 35 This reaction proceeds through a reversible coordination of the thiol to the Au(III) center of SM, generating Int1. Reductive elimination (TSRE) from Int1 irreversibly forms the C(sp 2 )-S bond (Int2), then product dissociation forms the free S-arylated product (PAr) in addition to the Au(I) chloride starting material (PAu, Fig. 2A). 41 These calculations revealed that reagents 5-7 had a low ∆GRE ‡ and a thermodynamically favorable ∆GRxn. Additionally, by calculating the %VBur of the Au(III) OAC (SM) of 5-7, %VBur of 70.0, 69.8, and 67.1 were obtained, respectively. 42 This steric parameter is in agreement with the experimentally determined kCo for each of these complexes wherein the more sterically available complexes (i.e. less %VBur) exhibited faster thiol coordination. We therefore hypothesized that the %VBur parameter could be used as a predictive metric for the rate of thiol coordination to the Au(III) center.
To further enhance this rate of coordination, we sought to screen smaller dialkylphosphines to provide more sterically available P,Nligated Au(III) OACs using NMe2 as the N-substituent. Accordingly, diisopropyl-(P i Pr2) (8) and dicyclopentylphosphine-based (PCp2) (9) P,N-ligated Au(III) SM geometries were calculated at the ωB97X-D/6-311+G(d,p), SDD, CPCM(Water)//B3LYP-D3/6-31G(d), LANL2DZ, CPCM(Water) level of theory. Complexes 8 and 9 had %VBur of 66.8 and 66.5, respectively. These values are very similar to the NMe2-PCy2-ligated OAC 7, so these ligands should affect a modest improvement on kCo. For 8 and 9, the comparable angles of the dialkylphosphine alkyl groups lead to similarly sized substituents compared to the cyclohexyl ring, which contributes to the small difference in %VBur compared to 7. To this end, smaller ring sizes on the dialkylphosphine fragment of the P,N-ligand were examined to affect larger changes in %VBur. Dicyclobutyl-(PCyb2) (10) and dicylopropylphosphine-based (PCyp2) (11) P,N-ligated OACs were determined to have 65.6 and 63.2 %VBur, respectively. These values demonstrate that the smaller alkyl substituents induce more steric availability around the Au(III) center (Fig. 2B). To determine the lower limit of %VBur for these dialkylphosphine P,N-ligated Au(III) OACs, dimethylphosphine-based (PMe2) P,N-ligated SM (12) was examined and found to have 61.6 %VBur.
Inspired by recent work developing bicyclo[1.1.1]pentylcontaining phosphines that are more strained and consequently smaller compared to their cycloalkyl counterparts, 43,44 the %VBur for the di(bicyclo[1.1.1]pentyl)phosphine-(PBcp2) (13) and di(bicyclo[1.1.0]butyl)phosphine-based (PBcb2) ( 14) P,N-ligated OACs were computed. Compound 13 had a %VBur of 65.5 while 14 had a %VBur of 63.5. These dialkylphosphines provide a significant decrease in %VBur compared to P,N-ligated SM 9 (PCp2) and 10 (PCyb2), which is in agreement with the additional strain in the bicyclic systems leading to smaller ring sizes and therefore less %VBur. The dicubylphosphine-based (PCubane2) P,N-ligated SM (15) was also examined, which could be employed due to its similarity to 6 (P t Bu2). Complex 15 had a %VBur of 65.9, indicating that formation of the cyclic scaffold does provide more steric availability to the SM. Taken together, these eight new Au(III) OACs (8-15) provide significant changes in %VBur compared to previously examined reagents 5-7 (Fig. 2D).
We previously determined that while the coordination to the Au(III) center is important, the subsequent, irreversible reductive elimination to form the C(sp 2 )-S bond (PAr) remains a key step in the S-arylation. 35 Accordingly, the ∆GRE ‡ and ∆GRxn were computed for each of the complexes. All of the reactions were calculated to be exergonic and irreversible. With the exception of the PMe2 Psubstituted OAC (12), the ∆GRE ‡ values are reasonable to undergo sufficiently fast reductive elimination at 25 °C (20.5-23.1 kcal mol -1 ), but the ∆GRE ‡ of 24.8 kcal mol -1 for 12 likely prevents it from performing an efficacious reductive elimination at ambient temperature (Fig. 2C).
While prior work from our group has focused on modifying the dialkylphosphine component of the P,N-ligand, there has been research varying the amino fragment of these P,N-ligands in Pd catalysis, most notably to morpholine and piperidine. [45][46][47][48][49] Accordingly, these Au(III) complexes with morpholine ( 16) and piperidine (17) were calculated with a PAd2 P-substituent (Fig. 3A),
Please do not adjust margins but there was a marginal difference in %VBur compared to the dimethylamino reagent (5, Fig. 3B). We hypothesized that by contracting the ring size, one could generate more sterically available Au(III) OACs. Pyrrolidine (18), azetidine (19), and aziridine (20) were consequently incorporated into the PAd2-based P,N-ligated Au(III) complex and determined to have %VBur of 69.5, 68.6, and 66.7 (Fig. 3B-D), respectively. These results are in line with the dialkylphosphine calculations (Fig. 2) wherein the five-membered ring is similar to the six-membered ring, but significant changes are observed in the four-and three-membered rings. To ensure that these amine modifications did not have significant effects on the thermodynamics of the reaction, the ∆GRE ‡ and ∆GRxn were computed for 16-20. Each reaction was calculated to be exergonic and irreversible with a low ∆GRE ‡ (12.2-18.8 kcal mol -1 ) (Fig. 3B). Notably, the ∆GRE ‡ for the piperidine-containing OAC ( 16) is significantly lower than the other complexes, potentially due to relaxation of the amine fragment from the twist-boat conformation in the S-coordinated Int1 to the chair conformation in the transition state (See ESI for details). This complex may warrant further examination to determine its ability to mediate other, more challenging reductive eliminations, 50 but the high %VBur indicates that it will likely not have a faster coordination than the other complexes examined herein.
To validate these computational results, we sought to synthesize the compounds with both a lower %VBur than the NMe2-PCy2 ligated OAC (7) and a reasonable ∆GRE ‡ . Accordingly, the PCp2-, PCyb2-, and PCyp2-based P,N-ligands with NMe2 as the N-substituent were targeted due to their predicted rapid coordination and sufficiently fast reductive elimination. While the bicyclic systems 13 and 14 did have lower %VBur compared to their cyclic counterparts, they each had a similar %VBur to the more synthetically accessible PCyb2- (10) and PCyp2-based (11) systems. Due to this and their synthetic difficulty, 13 and 14 were not further examined. For the amine components, we wanted to synthesize the PAd2 ligand with azetidine and aziridine as the amino substituent. However, the phenyl aziridine
Please do not adjust margins Please do not adjust margins complex polymerizes in water, 51 which makes it an unsuitable reagent to perform bioconjugation reactions. We also wanted to examine if modifying both the Pand the N-substituents of the ligand simultaneously generated an additive effect of the two substitutions. To do this, the azetidine-PCy2 complex was targeted.
The NMe2-PCp2 ligand 21 was synthesized by in situ formation of the aryl lithiate from 2-iodo-ortho-dimethylaniline and n-BuLi then substitution with the commercially available chlorodicyclopentylphosphine (See ESI for details). Incomplete conversion to the desired P,N-ligand necessitated purification of the ligand by preparative high performance liquid chromatography (HPLC). Attempts to generate the NMe2-PCyb2 and NMe2-PCyp2 ligands relied on formation of the aryl lithiate and in situ generation of the chlorodialkylphosphine using phosphorus trichloride and the corresponding Grignard reagent. Purification of these reagents by preparative HPLC induced significant formation of the phosphine oxide, limiting the utility of these ligands.
However, the azetidine-PAd2 ligand ( 22) and azetidine-PCy2 ligand ( 23) could be readily accessed by Pd-catalyzed C(sp 2 )-P bond formation from the corresponding ortho-halo-substituted aniline species and the dialkylphosphine in good isolated yield (80% and 63% yield, respectively). With P,N-ligands 21-23 isolated (Fig. 4A), the Au(I) complexes (24-26) were generated using HAuCl4•3(H2O) and 2,2'-thiodiethanol as a water-soluble variant of the commercially available Au(I)Cl(SMe2) (Fig. 4B, see ESI for details). The Au(III) OACs with a p-toluene aryl substituent (27-29) were synthesized using AgSbF6 as a halide scavenger and 4-iodotoluene to generate a model compound for kinetic investigations (Fig. 4C, see ESI for details).
With the Au(III) OACs synthesized, we wanted to test the kinetics of the thiol coordination to validate our computational predictions. To perform the coordination kinetics studies, stopped-flow/UV-vis spectroscopy experiments were performed at 5 µM of OAC with glutathione (GSH) as a model thiol under pseudo-first order conditions (500 µM -4 mM of GSH) at pH 7.4 and 20 °C (Fig. 5A, see ESI for details). By monitoring the absorbance, one can determine the formation of Int1 and extrapolate the observed second-order rate constant, kCo, from the experimentally observed rate constant (kObs). Examining 27, a kCo of 18,700 M -1 s -1 was calculated, which is faster than the NMe2-PCy2-ligated OAC 3 (16,600 M -1 s -1 ). Also, the equilibrium of this complex is significantly shifted toward the Scoordinated intermediate (Int1) with a KEq (kCo/kDis) of 1,530 M -1 (Fig. 5B-C).
Next, rate measurements were performed for 28 to determine if modifying the amine fragment was another viable strategy to affect the kCo. Kinetics experiments show that by utilizing the azetidine substituent, the kCo increases to 11,600 M -1 s -1 , which is nearly a fivefold increase compared to the NMe2-PAd2-ligated reagent 1 (2,460 M -1 s -1 ). By adding the Pand N-modifications together, we propose that 29 would provide the fastest kCo reported. The studies were performed and a kCo of 20,200 M -1 s -1 was observed for 29 (Fig. 5B-C), which represents a 25% increase in kCo compared to 3 and is the fastest observed rate for S-coordination to an organometallic Au(III) compound to date. These results are all consistent with the %VBur calculations (Fig. 2C and Fig. 3B), validating the use of %VBur as a steric descriptor to predict the rate of bimolecular coordination for these Au(III) OACs.
With the kCo determined for these three OACs, as well as the equilibria showing preferential thermodynamic formation of Int1, we next wanted to determine if the bimolecular coordination elementary step was still the selectivity-determining step in the reaction by performing S-arylation competition experiments. These experiments were performed by incubating one equivalent of GSH with three equivalents of two different OACs. These OACs vary by one methylene unit, which allows us to determine by liquid Please do not adjust margins Please do not adjust margins chromatography-mass spectrometry (LC-MS) which Au(III) reagent mediates the S-arylation more effectively (Fig. 6A). The NMe2-PCy2- (30) and NMe2-PAd2-ligated ( 31) Au(III) OACs with a p-ethylbenzene aryl substituent were prepared accordingly.
A competition experiment was performed between the NMe2-PCp2 ( 27) and NMe2-PCy2 ( 30) Au(III) OACs, which resulted in a 57:43 preference for the p-tolyl S-arylated product (32) over the pethylbenzene S-arylated product (33) (Fig. 6B-C). This preferential S-arylation from the NMe2-PCp2 OAC is in agreement with both the %VBur calculations as well as the kCo obtained from stopped-flow kinetics. To examine the amine fragments, competition experiments were performed using the PAd2-based P,N-ligated Au(III) OACs with azetidine (28) and dimethylamino (31) N-substitution. A 78:22 formation of 32:33 shows a preference for S-arylation from the azetidine-containing ligand (Fig. 6B-C), which is also in agreement with the %VBur calculations and the stopped-flow kinetics.
Please do not adjust margins However, when the competition experiments were performed using the PCy2-based P,N-ligated Au(III) OACs using the azetidine (29) and dimethylamino (30) N-substituents, a 32:68 ratio for 32:33 was observed (Fig. 6B-C). This indicates preferential S-arylation from the dimethylamino-containing ligand 30. The %VBur calculations and the stopped-flow kinetics indicate that coordination is faster to 29, but we hypothesize that the reductive elimination is slower, allowing for a reversible coordination. This leads to a potential switch in the selectivity-determining step for this reaction and indicates that while the two ligand permutations are additive for the thiol coordination event, both elementary steps of the S-arylation need to be sufficiently fast to induce effective conjugation.
To examine the effects of the azetidine compounds on the reductive elimination, stopped-flow/UV-vis spectroscopy was performed using the conditions described vide supra (Fig. 7A, see ESI for details). From Int1 of 28, a unimolecular rate constant was observed for the reductive elimination (kRE) of 2.66 s -1 (Fig. 7B), which is approximately three times slower than kRE of the NMe2-PAd2-ligated OAC 1 (Fig. 7D). This is also in line with the calculated ∆GRE ‡ for the phenyl OACs (Fig. 3B). Profiling the reductive elimination from Int1 of 29, a kRE of 0.133 s -1 was calculated (Fig. 7C), which is nearly ten times slower than the kRE for the NMe2-PCy2-ligated OAC 3 (Fig. 7D). These results confirm the proposed switch in selectivity-determining step for 29 as observed in the competition experiments. Altogether, this work demonstrates that %VBur can serve as a quickly calculated metric to screen the rate of bioconjugation to an Au(III) OAC, which will be valuable for the future screening of Au(III) and other metal complexes for bioconjugation. While the %VBur can be used to predict the coordination to the Au(III) center, it can also be used to predict the rate of reductive elimination, though the trend is in the opposite direction (Fig. 8). We propose that this workflow should be amenable to other metal-mediated S-arylation reactions and provide a rapid way to screen these reagents. This work also provides fundamental mechanistic insight into the elementary steps of Au(III) chemistry, 36,[52][53][54][55][56][57] which is relatively understudied compared to Pd(II) and Pt(II) systems.
Thirteen Au(III) OACs were examined with eight different dialkylphosphines and five different amines in silico to determine their effects on both steps of the S-arylation reaction. Three novel Au(III) OACs were synthesized (27-29) and their bimolecular rate constants were determined to be between 11,600-20,200 M -1 s -1 . Finally, these OACs were examined in competition experiments to elucidate which of the two elementary steps of the S-arylation controls the selectivity, and a switch in the selectivity-determining step was confirmed experimentally for one of these reagents.