Selective Iridium-Catalyzed Reductive Amination Inside Living Cells

Abstract

Given that amino groups are ubiquitous in bioactive molecules, abiotic routes to incorporate them into cellular species offer new opportunities to study and manipulate living systems. In the present work, we report the first biocompatible method to prepare 1°, 2°, or 3° amines selectively starting from an aldehyde and nitrogen precursor through iridium-catalyzed reductive amination. To prevent overalkylation, we developed a non-toxic self-immolative agent comprising 4-(1-aminoethyl)phenol that can condense with carbonyl groups and undergo 1,6-elimination upon reduction to the desired 1° amines. The use of an electron-poor half-sandwich Ir catalyst favored the formation of amine over alcohol products. To synthesize 2° or 3° amines, the aldehydes were combined with the appropriate 1° or 2° amine, respectively, under our standard reaction conditions. Our method is sufficiently mild to perform on proteins, as demonstrated by the conversion of aldehyde-containing allysine residues in bovine serum albumin to lysine. Importantly, we showed that Ir-catalyzed reductive amination could be applied inside living cells, such as by generating the alkaloid phenethylamine or calcium-reducing drug cinacalcet to elicit different biological responses. The amines formed via intracellular reductive amination were quantified by high performance liquid chromatography, revealing that turnover numbers of up to ~20 were achieved. This work is expected to enable greater versatility and precision in transforming a wide range of aldehyde-containing entities within living environments, further expanding our biosynthetic chemistry toolbox.

Graphical Abstract

Introduction

The amino group is a key feature of numerous biomolecules and pharmaceuticals,^1–5 owing to its Lewis basicity, nucleophilicity, hydrogen bonding capability, and/or tetrahedral geometry. Given the importance of amines in biology, cellular processes that lead to unregulated deamination can be detrimental to human health.^6,7 For example, the breakdown of dopamine in the brain leads to the formation of 3,4-dihydroxyphenylacetaldehyde, which is neurotoxic and has been strongly implicated in Parkinson’s disease.^8,9 The oxidative deamination of serotonin, another important signaling molecule in the central nervous system, to 5-hydroxyindoleacetaldehyde may be involved in protein oligomerization linked to neurodegeneration.^10,11 Thus, biosynthetic tools that can repair deaminated species in living systems may enable new therapeutic advances. Another potential application of intracellular amine synthesis is the in-cell generation of medicinal compounds.^4,12,13 Some examples of amine-containing drug molecules approved by the United States Food and Drug Administration (FDA) include the anticonvulsant pregabalin (1° amine),¹⁴ the stimulant lisdexamfetamine (1° amine),^15,16 and calcium reducer cinacalcet (Cin, 2° amine).¹⁷ In some cases, generating the bioactive species in situ may be more beneficial than administering the active drugs, especially when the drug has a short shelf life or poor uptake.

A possible strategy to convert carbonyls to amines is through reductive amination (RA).¹⁸ This process involves condensing an aldehyde or ketone with an appropriate nitrogen donor to generate the corresponding imine, followed by reduction to obtain the target amine. However, to achieve efficient RA, several challenges must be overcome, such as overalkylation, competing reduction of carbonyls to alcohols, toxicity of the nitrogen donors, and/or instability of metal catalysts in biological media.^19–23 Conventional RA methods include catalytic hydrogenation^12,24–26 or stoichiometric hydride reduction^27,28 (Chart 1A). These approaches are used widely in chemical synthesis but are incompatible with living systems due to the need for cell impermeable heterogeneous catalysts or potentially toxic hydride sources.²⁹ For example, it was reported that NaBH3CN was used with cells, but the specimen in this study were exposed to the fixative agent formaldehyde prior to being treated with the reducing agent so the cells are likely non-living.³⁰ The use of ammonia to produce 1° amines in living cells is a non-starter because NH3 is a potent cell growth inhibitor.^31–33

Chart 1.

Reductive amination of aldehydes to amines using conventional (A) and transfer hydrogenation-based methods (B). We have applied our method in living cells to synthesize phenethylamine and cinacalcet (C). R = alkyl or aryl; R′, R′′ = H, alkyl, or aryl; X′ = self-immolative group; Ir_A = Cp* iridium catalyst with bipyridine ligand, Ir_B = Cp* iridium catalyst with picolinamidate ligand.

Another synthetic approach to convert carbonyls to amines relies on transfer hydrogenation utilizing formate salts in combination with an organometallic catalyst.³⁴ This method has been applied successfully to prepare 1° amines but the reactions are typically conducted in organic solvents with strong acids and/or high temperatures (up to 80 °C),^35–38 conditions that are unsuitable for mammalian cells. The application of ammonium formate, which can be a source of ammonia depending on the solution pH, could potentially pose toxicity concerns.³³ A notable example of transfer hydrogenative RA was reported by McFarland and Francis (Chart 1B),³⁹ who demonstrated that the lysine residues in lysozymes can be alkylated by treating the proteins with an aldehyde, sodium formate, and Ir complex in phosphate buffer at pH 7.4 and 22 °C. However, the small-molecule amine synthesis scope and selectivity as well as catalytic efficiency of this method were not reported. Furthermore, no studies in living cells were performed.

In efforts to bring new chemistry to life, researchers have developed abiotic catalysts based on molecular^19–23 or nanoparticle^40,41 scaffolds that can promote synthetic reactions in living cells. A few notable examples include azide-alkyne cycloaddition,^42,43 protecting group cleavage,^44–49 olefin metathesis,⁵⁰ and others.⁵¹ In particular, the demonstration that transition metal complexes⁵² (e.g., those containing Ir,^53–56 Ru,^57–60 or Os^61,62) could catalyze the transfer hydrogenation between hydride donors and acceptors in aqueous environments or cells suggest that it might be possible to employ such catalyst systems in an intracellular RA scheme.

Herein, we report the first biocompatible method to convert aldehydes selectively to 1°, 2°, or 3° amines via transfer hydrogenative RA (Chart 1B). This innovation was made possible by the development of a self-immolative nitrogen donor that allowed the controlled release of 1° amines and the use of electron-deficient organoiridium catalysts favoring the formation of amine over alcohol products. Our method is efficient for a wide variety of substrates and is mild enough for applications on proteins. We showed that our Ir complexes promote RA inside living cells by converting a potentially toxic metabolite phenylacetaldehyde into the less toxic alkaloid phenethylamine,^63–65 leading to enhancement in cell viability (Chart 1C). In contrast, in-cell synthesis of cinacalcet via abiotic RA led to diminished human neuroblastoma cell growth due to the drug’s inhibiting effects.^17,66,67 Quantification of the intracellular reactions revealed that up to 20 turnovers were achieved, indicating that the Ir complexes can regulate biological systems through their catalytic action, similar to natural enzymes.

Results and Discussion

Evaluation of Product Selectivity.

Because conventional methods suffers from multiple biocompatibility issues (Chart 1B), we focused our efforts on approaches involving metal-catalyzed transfer hydrogenation.^{54,59,62,68,69} Our survey of the synthetic chemistry literature revealed that methods employing ruthenium³⁶ or iridium^{35,37,38,70–72} catalysts for RA of carbonyls to 1° amines typically require ammonia, formic acid, organic solvent, and/or high temperature (>37 °C). However, our previous studies using a [Cp*Ir(picolinamidate)Cl] catalyst (where Cp* = pentamethylcyclopentadiene anion, Ir1; Figure 1A) suggest that performing RA under physiologically relevant conditions might be possible.⁷³ However, only one example involving the conversion of sodium pyruvate to alanine in aqueous media was demonstrated successfully in this work. Given that Ir1 can catalyze the reduction of aldehydes to alcohols inside living cells,^54,55,74,75 we hypothesized that further developments using this catalyst system might enable intracellular RA.⁷⁶

Figure 1.

Structures of Ir complexes used in this study (A) and comparison of products obtained from the reaction of 1a with Ir and HCOONH₄ (B). The reaction products were quantified by GC using pentamethylbenzene as an internal standard. Products not detected are marked with an asterisk (*) above the corresponding column. Yield % = [(moles of product formed × reaction stoichiometry/moles of starting aldehyde) × 100].

To screen for RA activity in biomimetic environments, we combined hydrocinnamaldehyde (1a), HCOONH4, and an Ir complex (1 mol% relative to the aldehyde) in DMSO/phosphate buffered saline (PBS) (5:95) at 37 °C in air for 20 h (Figure 1B, Table S1). The Ir catalysts were prepared as described previously or using a similar procedure (see SI for details).^77,78 The reaction products were extracted into ethyl acetate and then quantified using gas chromatography (GC). Our results showed that the [Cp*Ir(4-R-picolinamidate)Cl] catalysts (where R = H or a functional group) afforded different product distributions. The parent Ir1 (R = H) gave mostly bis(3-phenylpropyl)amine (4a, 20%) and tris(3-phenylpropyl)amine (5a, 63%) products, with no 3-phenylpropylamine (3a) and minor amounts of 3-phenylpropan-1-ol (2a, 9%). Catalysts with electron-donating or chloro substituents, such as Ir2 (R = NMe₂), Ir3 (R = OMe), Ir4 (R = OH), or Ir5 (R = Cl), also produced a mixture of alcohol and 2°/3° amine products. The low overall yield obtained using Ir4 may be due to the non-innocence of the ligand’s hydroxy group, which could potentially bind and inhibit another Ir species. In contrast to the electron-rich catalysts, the electron-poor variants, Ir6 (R = CF₃), Ir7 (R = CN), and Ir8 (NO₂), provided only 2°/3° amines in >90% yield, with negligible quantities of the alcohol species. Surprisingly, [Cp*Ir(4,4′-di-methoxy-2,2′-bipyridine)Cl] (Ir9),³⁹ the catalyst used by McFarland and Francis for RA of lysine residues in lysozymes, was inefficient under the conditions tested, giving only a 10% combined yield of 4a and 5a. This observation is consistent with our previous finding that Cp*Ir complexes with 2,2′-bipyridine ligands can react with formate to generate Ir-H intermediates but do not readily transfer its hydride to organic acceptors.⁷⁸

The results above indicate that accessing primary amines (e.g., 3a) from RA of aldehydes is highly challenging due to several competing processes. As shown in Scheme 1A, an aldehyde substrate I can undergo reduction by an Ir-H species to the corresponding alcohol II or condense with NH₃, which could be derived from ammonium salts, to afford imine III. Based on our previous studies,⁷⁸ electron-rich Ir-H species will transfer its hydride to I more rapidly than its electron-poor counterpart. Thus, to favor imine condensation (I→III) over direct aldehyde reduction (I→II), the Ir catalyst should be electron-deficient. The underlying reason for this selectivity is analogous to that for choosing milder (e.g., NaBH₃CN or NaBH(OAc)₃) over stronger (e.g., NaBH₄) hydride reducing agents to facilitate one-pot RA.¹⁸ Once III is formed, reaction with an Ir-H intermediate would produce the corresponding 1° amine IV. Unless a large excess of ammonia is present to react with all of the starting aldehyde, the amine products can proceed through subsequent imine condensation and hydride transfer reactions to generate 2° (IV→V→VI) and 3° (VI→VII or VIII→IX) amines. Therefore, additional strategies are needed to achieve selectivity in RA, especially to obtain 1° amines from aldehydes.

Scheme 1.

Possible routes for reductive amination using Ir catalysts with HCOONH₄ (A) or HCOONa and a self-immolative nitrogen donor (B). In Part B, no secondary or tertiary amines were observed. R = alkyl or aryl, X′ = self-immolative group.

Synthesis of 1° Amines.

To prevent the overalkylation of 1° amine products, we focused on developing self-immolative nitrogen donors that are biocompatible and can serve as blocking groups to hinder the formation of higher order species. This strategy was inspired in part by work reported by Kuwata and coworkers,^71,72 which showed that 1° amines can be prepared via RA using β-amino alcohols as a degradable aminating agent in a two-step process. However, this method requires acid and an external oxidant, which are not desirable for biological applications. In our proposed strategy, the self-immolative donor NH₂X′ (where X′ = a self-cleavable group) can condense with I to afford the corresponding imine X (Scheme 1B). Reaction of X with an Ir–H species would provide XI, which would then disassemble spontaneously to the desired 1° amine under suitable conditions. Furthermore, we postulated that the use of an electron-poor iridium catalyst would minimize the reduction of aldehyde to alcohol (i.e., I→II).

In our design of self-immolative nitrogen donors, we favored structures comprising 4-hydroxybenzylamines or their precursors so that upon alkylation via RA, its X′ group can eliminate under physiological pH or in the presence of oxidants to release the desired 1° amines (Scheme 1B).^79,80 To screen the candidates, the commercially available NH2X′ were combined with 1a, HCOONa, and Ir6 in DMSO/PBS (5:95) under air at 37 °C for 24 h (Figure 2A). GC-MS analysis of the products showed that reactions using the boronic acid (6) and boronic ester (7) derivatives gave low yields of 3a (≤15 %), which was not surprising given that the boronic acid/ester groups would likely need to be oxidized to a hydroxy group for 1,6-elimination to occur.^81–83 However, when we tested 4-hydroxybenzylamine (8) and 4-hydroxy-3-methoxybenzylamine (9) as the aminating agent, only 18% and 26% yields of 3a were obtained, respectively. Increasing the steric bulk of the nitrogen donor using 4-(1-aminoethyl)phenol (10) gave a substantially higher amount of the primary amine, with up to 93% of 3a. The O-alkylated 4-methoxy-α-methylbenzylamine (11) provided the desired amine in only 21% yield, indicating that the free hydroxyl group in the benzylamine structure is necessary for efficient self-immolation.

Figure 2.

Evaluation of nitrogen-precursors as possible self-immolative donors (A) and comparison of products obtained from the reaction of 1a with Ir, HCOONa, and 10 (B). In Part B, no secondary (4a) and tertiary (5a) products were obtained. The reaction products were derivatized by treatment with ethyl chloroformate and then quantified by gas chromatography using an internal standard. Products not detected are marked with an asterisk (*) above the corresponding column. Yield % = [(moles of product formed × reaction stoichiometry/moles of starting aldehyde) × 100].

Encouraged by these results, we evaluated RA selectivity using 1a, 10, and HCOONa in the presence of different Ir catalysts under our optimized reaction conditions. Unlike in our experiments using ammonium salts as the nitrogen source (Figure 1B), reactions in the presence of 10 provided only 3a without any overalkylation products 4a and 5a (Figure 2B). Our data showed that the electron-poor catalysts Ir6, Ir7, and Ir8 produced the desired primary amine with only trace amounts of alcohol 2a. Catalyst Ir6 was the most selective, giving a 3a:2a ratio of ~133 (Table S3). In contrast, the parent (Ir1), electron-rich (Ir2, Ir3), and chloro-bearing (Ir5) catalysts had a much lower 3a:2a ratio of ≤4.5. Reactions performed using Ir6 in commercial cell culture media Dulbecco’s Modified Eagle Medium (DMEM) also gave primarily 3a (94% yield), although the overall yield was lower in the presence of high concentrations of biological additives (e.g., 10 mM of lysine or 500 μM of glutathione; Figure S7) due to their catalyst inhibitory effects or competing reactivity (e.g., condensation of 1a with the lysine side chain instead of with 10). Solution studies by NMR spectroscopy have shown that glutathione and other nucleophilic biomolecules can readily bind half-sandwich Ir catalysts.^84–86 It should be noted that in all reactions in which primary amines were produced, the cleaved auxiliary 4-ethylphenol (12) was also observed. Furthermore, the reactions all appeared to be homogeneous. These results provide strong support that achieving high selectivity for 1° amine products in RA requires both a bulky self-immolative nitrogen donor and an electron-deficient Ir catalyst.

Having established a robust RA protocol using Ir6, we next studied the aldehyde scope of our method (Figure 3, Table S5). We observed that araliphatic aldehydes (1a-1c) were converted to the corresponding primary amines (3a-3c) with yields between 86–93%. Cinnamaldehyde (1d), which is an α,β-unsaturated aldehyde, was aminated to a primary amine but the C=C bond was also reduced to afford 3a in 88% yield. The aliphatic aldehydes n-octanal (1e), n-hexanal (1f), n-butanal (1g), and 3-methylbutanal (1h), were successfully transformed to the desired primary amines 3e, 3f, 3g, and 3h, respectively, in moderate to high yields (56–85%). While benzaldehyde (1i) was readily converted to both benzyl alcohol (2i, 40% yield) and benzylamine (3i, 48%), the electron-rich derivative 4-dimethyl-aminobenzaldehyde was significantly less reactive, giving the corresponding amine 3j in only 23% yield. The natural compound vanillin (1k),^87,88 which has been shown to possess anti-inflammatory, anti-cancer, and anti-depressant properties, was unreactive under our experimental conditions. Similarly, we were unable to aminate the ketone acetophenone 1l. From a biological standpoint, the different reactivity of carbonyl-containing substrates in Ir-catalyzed RA is useful because it may allow favorable discrimination of various aldehydes inside living cells. In particular, toxic aldehydes are generally linear and α,β-unsaturated whereas essential aldehydes are often bulky and aromatic (e.g., vanillin).^89–93 Thus, given that Ir6 appears to react more readily with the former than the latter, it may be promising as an intracellular catalyst for RA.

Figure 3.

Reductive amination of various aldehydes to the corresponding primary amines using Ir6, HCOONa, and 10. The reaction products were derivatized by treatment with ethyl chloroformate and then quantified by gas chromatography using an internal standard. Unless indicated (e.g., 2i, 2k), no alcohol products were observed. Yield % = [(moles of product formed × reaction stoichiometry/moles of starting aldehyde) × 100]. See Table S5 for more details.

Mechanism of Selective Reductive Amination.

Our proposed mechanism for the RA of 1a in the presence of 10 and Ir catalyst is shown in Figure 4A. This process likely involves imine condensation between 1a and 10 to yield 3a_imine, which can then be reduced to 3a_pro via hydride transfer from an Ir-H intermediate generated from an iridium catalyst and HCOONa. Theoretical studies by Liu, Cheng, and coworkers on related Ir-catalyzed RA provide computational support for such as a reaction sequence.⁹⁴ We hypothesize that 3a_pro is too sterically bulky to condense with another aldehyde, which is key to preventing overalkylation. The cleavage of 3a_pro could potentially occur through either direct 1,6-elimination or protonation, followed by 1,6-elimination to 3a and the cleaved auxiliary 12_quin.⁷⁹ Because 12_quin is electrophilic, it can react with an Ir-H species to generate 12. Thus, the presence of 12 would indicate that successful RA reactions have occurred.

Figure 4.

Proposed pathway for the conversion of 1a to 3a (A), kinetic studies of reductive amination and amine cleavage (B), and effects of pH and H₂O₂ on the amine cleavage efficiency (C). The kinetic studies were performed in DMSO/PBS (5:95) at 37 °C and the products were analyzed by HPLC (reductive amination) or GC (amine cleavage). [A_o] = concentration of starting material, [B] = concentration of product.

To obtain experimental support for the mechanism above, we followed the reaction progress by analyzing aliquots of the crude mixture at different times using high-performance liquid chromatography (HPLC) and GC. We observed that the reaction of 1a using 10, HCOONa, and Ir catalyst under our standard reaction conditions produced 3a_pro within the first ~90 mins (Figure S8). The data from the initial time points (0–40 min) appeared to follow pseudo-first order kinetics, giving rate constants k_obs1 of 11.4×10⁻⁴ s⁻¹ for Ir1 and 8.3×10⁻⁴ s⁻¹ for Ir6 (Figure 4B, left). Assuming that the imine condensation of 1a and 10 is independent of the Ir catalyst, the slight differences in rates between Ir1 and Ir6 may be due to faster hydride transfer by Ir1-H to the imine intermediate 3a_imine relative to that by Ir6-H. The rate of imine condensation and hydride transfer is decreased when the nitrogen donor does not bear a hydroxy group (e.g., using α-methylbenzylamine instead of 10, Figure S17) and the amine product does not undergo C-N bond cleavage. Beyond 90 min, we observed the disappearance of 3a_pro concomitant with the appearance of 3a (Figure S12). The initial rates of 3a formation were estimated using pseudo-first order kinetics, giving k_obs2 of 4.2×10⁻⁵ s⁻¹ for Ir1 and 4.8×10⁻⁵ s⁻¹ for Ir6 (Figure 4B, right). These results suggest that amine cleavage is not dependent on the Ir catalyst and that it is the rate-limiting step in the overall RA process.

To investigate the amine cleavage step further, we independently synthesized 3a_pro (see SI for details) and studied its self-immolation efficiency under different conditions. We observed that the conversion of 3a_pro to 3a is highly sensitive to pH, in which maximum release of the primary amine occurred between pH 7.4 and 9.0 at 37 °C (up to 70% yield) (Figure 4C, left). Within this pH range, the amine group in 3a_pro is expected to be protonated since the pK_a value of PhCH₂NH₃⁺ in water is ~10.^95–98 Thus, protonation of 3a_pro to 3a_proH, followed by 1,6-elimination of the 4-ethylphenol group is likely the preferred route of amine cleavage (Figure 4A). At this time, we are uncertain whether this process is concerted or stepwise (i.e., if carbocation intermediates are formed). Density functional theory calculations on related systems suggest that deprotonation of the OH group of the ammonium intermediate initiates the dissemble sequence.^99,100 The addition of Ir1 or Ir6 to 3a_pro at pH 7.4 did not improve the yield of 3a, indicating that the Ir catalysts do not play a role in the C-N bond scission (Table S6).

When 3a_pro was treated with increasing amounts of H₂O₂, the formation of 3a increased slightly (Figure 4C and Table S7). For example, in the presence of 250 mM of H₂O₂ at pH 7.4, 92% yield of 3a was obtained compared to 67% in the absence of H₂O₂. These results indicate that oxidation of 3a_pro to 3a and 12_quin is another possible route to amine disassembly. This possibility was further supported by kinetic studies of RA, showing that the amine cleavage step is accelerated in the presence of air relative to nitrogen (Figure S16). Given that O₂ in air can react with Ir-H species in water to generate H₂O₂,^53,101 oxidative cleavage could be a minor competing pathway for amine release.

Synthesis of 2° and 3° Amines.

Although notable advances have been made in transfer hydrogenative RA of carbonyl compounds to 2° and 3° amines,^102–110 most methods reported are not biocompatible (e.g., require low pH, organic solvents, and/or high temperatures). To prepare higher order amines through Ir catalysis, we studied reactions using 1a (5 mM), benzylamine (7.5 mM), HCOONa (50 mM), and various Ir complexes (50 μM) in DMSO/PBS (5:95) at 37 °C for 24 h (Table S9). We found that the majority of catalysts (Ir1, Ir3, Ir6, Ir7, and Ir8) gave N-(3-phenylpropyl)benzylamine (14) in significantly larger quantities than N,N,-(bis(3-phenylpropyl))benzylamine (15), with a 14:15 ratio of 90 with Ir7. In contrast, the dimethylamino-functionalized Ir2 and chloro-functionalized Ir5 had lower 14:15 ratios of 3 and 11, respectively. The bipyridine-ligated Ir9 afforded poor yields of 2° and 3° amines (18% total), indicating that it is not an efficient catalyst for amine synthesis under these conditions. When the amount of benzylamine was increased from 7.5 to 10 mM, greater selectivity toward 2° over 3° amines was achieved (14:15 = 195, Table S8).

Using our optimized reaction conditions, we evaluated the substrate scope in 2° amine synthesis. Reactions starting with 1a and various 1° amines in the presence of Ir6 and HCOONa produced the desired 2° amine as the major product in most cases (Figure 5A). Benzylamine, cyclohexylamine, and α-methylbenzylamine gave >90% yield of the 2° amines 14, 16, and 17, respectively. Unfunctionalized and electron-rich anilines were efficiently converted to the corresponding N-(3-phenylpropyl)aniline (19-21, 86–99% yield) with high selectivity. However, 4-nitroaniline was unreactive, suggesting that it is too electron deficient to react with 1a. A variety of araliphatic and alkyl amines, including 2-phenylethylamine, 3-phenylpropylamine, n-hexylamine, and n-butylamine, underwent RA with 1a to provide the desired amines 23, 4a, 25, and 27, respectively, in appreciable yields (63–89%). However, the 2°:3° ratios in these reactions were variable, ranging from 8 to 22. Finally, the bulky isobutylamine was not a suitable substrate, giving only 13% yield of the 2° amine 29 and 34% yield of the 3° amine 30. Reaction of 1a with an N-methylated variant of 10 under our RA conditions afforded a mixture of 2° and 3° amines (Figure S20), suggesting that the use of a self-immolative precursor did not improve product selectivity in the synthesis of higher order amines.

Figure 5.

Synthesis of secondary (A and B) and tertiary (C) amines using Ir6, HCOONa, and an appropriate nitrogen-containing starting material. The reaction yields were quantified using GC with pentamethylbenzene as an internal standard. Cinacalcet (part B) was obtained from the reaction of aldehyde 1p (1 mM), (R)-41 (2 mM), Ir6 (40 μM, 4 mol%), and HCOONa (10 mM) in DMSO/PBS (5:95) at 37 °C for 24 h. Yield % = [(moles of product formed × reaction stoichiometry/moles of starting aldehyde) × 100]. See Tables S10–S12 for more details.

To explore the aldehyde scope in 2° amine synthesis, different aldehydes were combined with benzylamine, HCOONa, and Ir6 under our standard conditions (Figure 5B and Table S11). Our results demonstrated that a variety of aldehydes could be aminated with exclusive selectivity for 2° over 3° amines. They include benzaldehyde to 31 (78% yield), 4-nitrobenzaldehyde to 33 (71% yield), furan-2-carboxaldehyde to 35 (83% yield), and n-hexanal to 36 (91% yield). The substrate cinnamaldehyde reacted with benzylamine but the resulting product was fully reduced, including the C=C double bond to furnish 14 in 82% yield. When acetophenone was employed as a substrate, no amine products were obtained, indicating that ketones are not compatible with our Ir-catalyzed RA method.

Given that the synthesis of 3° amines do not suffer from selectivity issues, we screened only a limited number of amine substrates (Figure 5C and Table S12). We found that reactions containing 1a, a 2° amine, HCOONa, and Ir6 in DMSO/PBS (5:95) produced the desired products in moderate to excellent yields (69–94%). Starting with methylbenzylamine, di-n-butylamine, or dimethylamine gave the corresponding tertiary amines 38, 39, or 40, respectively, with no alcohol side products.

To show the utility of our RA method, we employed it to prepare clinically important drug molecules. For example, the reaction of 3-(3-(trifluoromethyl)phenyl)propanal (1p) and (R)-1-(1-naphthyl)ethylamine ((R)-41) in the presence of HCOONa and Ir6 produced cinacalcet in 79% yield (Figure 5B). Alternatively, combining N-(3-phenylpropyl)ethylamine and 1a using the same catalyst system afforded 85% yield of the gastrointestinal drug Alverine (Alv) (Figure 5C).

Reductive Amination on Proteins.

To explore the feasibility of applying abiotic RA on biomacromolecules, we selected bovine serum albumin (BSA) as a model system because it is robust, readily available, and water-soluble. Given that allysine plays an important role in both normal and abnormal protein crosslinking,^111,112 we wanted to study its reactivity within a natural protein. To generate the allysine residues in BSA, we used a reported procedure by treating the native protein with ferric chloride, sodium ascorbate, and H₂O₂ in PBS and then stirring for 24 h (Figure 6).¹¹³ The desired BSA^Ald was purified using a desalting column and then characterized (vide infra). With the BSA^Ald protein in hand, we attempted to convert the aldehyde side chains to primary amines by combining the oxidized protein with 10, HCOONa, and Ir6 in DMSO/PBS (5:95) under air at 37 °C for 24 h. The resulting BSA^RA was purified by desalting and then subjected to further analysis.

Figure 6.

Oxidation of the lysine side chains in bovine serum albumin (BSA) to allysine (BSA^Ald), followed by reductive amination using Ir6 in the presence of 10 and HCOONa to produce BSA^RA. The bottom plots show the carbonyl density (left) and lysine content (right) in the protein samples after various treatments. The data were analyzed using one-way ANOVA and are presented as the mean ± standard deviation (n = 4 per group). The p-values are indicated as follows: ns = not significant (p > 0.05), * = p < 0.05, ** = p < 0.01, and *** = p < 0.001, **** = p < 0.0001.

Using commercial colorimetric assays, we determined the protein concentration, carbonyl density, and lysine content in the BSA samples after various treatments. We found that all of the protein samples had an average concentration of 19±2 mg/mL (Table S13), suggesting that the different manipulations did not cause significant protein degradation or aggregation. The BSA^Ald sample had a carbonyl density of 26.8 nmol/mg of protein, which is significantly higher than that of the native BSA (1.0 nmol/mg of protein) (Figure 6, bottom left). When BSA^Ald was treated with 10 only or 10/Ir6 (10 or 20 μM)/HCOONa, the aldehyde level was lowered to 4.4 and ≤3.2 nmol/mg of protein, respectively. Presumably, reaction of BSA^Ald with 10 formed imine adducts but could not be reduced to 1° amines in the absence of Ir6/HCOONa. This hypothesis is supported by our lysine measurements (Figure 6, bottom right). For example, BSA^Ald contained 15% amine compared to the native protein (100% amine). The BSA^Ald/10 sample had only 22% amine, which is consistent with the presence of imine rather than amine groups in the protein. In the BSA^Ald protein exposed to 10/Ir6/HCOONa, the lysine content increased from 52 to 79% as the Ir6 treatment concentration was increased from 4 to 20 μM. This Ir-dependence strongly suggests that transfer hydrogenative RA of allysine is responsible for converting the aldehyde to amino groups.

Reductive Amination in Living Cells: Eliciting Beneficial Effects.

To predict the Ir complexes’ ability to permeate cells, we determined their lipophilicity by measuring their partition between octanol and PBS using a shake-flask method.¹¹⁴ The partition coefficients (Log P) were determined to be 1.24, 1.46, 1.19, and 1.43 for Ir1, Ir3, Ir6, and Ir7, respectively (Figure S25). Given that these values are >1.0, the Ir complexes are considered to have lipophilic character, which tends to be favorable for cellular uptake.^115,116

To evaluate the cellular uptake efficiency, we incubated mammalian cells with 10 μM of an Ir complex for 24 h and then the cells were detached, lysed, and analyzed by inductively-coupled plasma mass spectrometry (ICP-MS). In NIH-3T3 mouse fibroblast cells, the Ir concentrations ranged from 54–136 ng/10⁶ cells in the order Ir3 < Ir7 < Ir1 < Ir6 (Table S14). In SH-SY5Y human neuroblastoma cells, the Ir accumulation was found to be between 68–169 ng/10⁶ cells, also in the order Ir3 < Ir7 < Ir1 < Ir6 (Table S15). These cell lines were chosen because they are commonly used in biological studies of metal complexes.^54,117 This cell uptake trend is negatively correlated with the Log P trend (Ir3 ~ Ir7 > Ir1 ~ Ir6; Figure S25). Although lipophilicity and cell uptake can sometimes have positive correlations,¹¹⁵ this is not always the case given that many factors can impact the entry of molecules into cells, such as their transport mechanism or interactions with various biological components.^118,119

To demonstrate our abiotic RA method in living cells, we investigated the conversion of phenylacetaldehyde (1c) to phenethylamine (3c) (Figure 7A). Because 1c has been implicated in neurodegeneration,⁶⁵ whereas 3c is a beneficial neuromodulator,⁶⁴ this transformation could have useful therapeutic potential. Two different procedures were tested to achieve optimal biological response. In our first attempt, we incubated cells with 1c (200 or 400 μM), 10 (1 mM), HCOONa (1 mM), and Ir6 (5, 10, or 20 μM) sequentially with washes in between each step (Figure S30). At the conclusion of the experiment, the cell viability across different treatment groups was assessed using a colorimetric sulforhodamine B assay. We found that 400 μM of 1c, which has a 50% inhibition concentration (IC₅₀) of 185 μM in NIH-3T3, was necessary to reduce cell viability to an appreciable extent. Unfortunately, the addition of the aminating agent 10, Ir6, and HCOONa to the aldehyde-treated cells did not result in enhanced cell growth. We hypothesized that the washing steps may have removed one or both of the starting reactants (i.e., 1c and 10), precluding productive reactions to generate 3c.

Figure 7.

A) Diagram showing treatment of NIH-3T3 mouse fibroblast cells with aldehyde 1c, followed by Ir-catalyzed reduction to the corresponding alcohol 2c or amine-containing natural compound 3c. At the conclusion of the experiment, the cells were lysed and the contents were derivatized with Fmoc-Cl prior to product analysis. B) Cell viability data obtained from carrying out reductive amination of 1c inside NIH-3T3 cells. C) HPLC analysis of products generated inside cells after treatment with 1c, 10, Ir1 or Ir6, and HCOONa. The turnover numbers (TONs) were calculated by dividing the product concentration by the intracellular iridium concentration.

In our modified procedure, we incubated the NIH-3T3 cells with 1c (400 μM) and HCOONa (1 mM) for 3 h, washed with fresh media, and then added 10 (1 mM) and Ir catalyst (10 μM) together for 24 h before measuring the cell viability (Figures 7B and S32). Our results showed that in the presence of 1c only, the cell viability dropped to 38% relative to that in the untreated control. Cells exposed to 10, Ir1, or Ir6 without aldehyde had viability of ≥84%, which is expected given that their concentrations were well below the toxic threshold (IC₅₀ = >8 mM, 62 μM, and 46 μM for 10, Ir1, Ir6, respectively). When the 1c-treated cells were combined with Ir1 or Ir6 and HCOONa, the cell viability increased from 38% to >70%. This enhancement is likely due to transfer hydrogenative reduction of 1c to the alcohol 2c or transfer hydrogenative RA to primary amine 3c, which are both non-toxic based on their IC₅₀ values of >7 mM. Similar cell enhancing effects were observed when Ir3 or Ir7 were used as the catalyst (Figure S32).

To obtain support for the intracellular reactions, we quantified the amounts of 2c and 3c produced inside NIH-3T3 cells treated with 1c, 10, Ir catalyst, and HCOONa. These experiments were performed in two separate 100 mm tissue culture plates, rather than 96-well plates, to generate enough product for chemical analysis. After the treatment period, the cells were combined and lysed, and the crude mixture was reacted with fluorenylmethyloxycarbonyl chloride (Fmoc-Cl) to derivatize 2c and 3c to the corresponding 2c′ and 3c′, respectively. The fluorenyl group in Fmoc is emissive so the derivatized compounds can be detected with higher sensitivity using HPLC equipped with a fluorescence detector. The HPLC traces in Figure 7C showed that 2c and 3c (as 2c′ and 3c′, respectively) were formed in the intracellular reactions. By dividing the product concentration by the Ir concentration in the cells, the turnover numbers (TONs) were determined (Table S17). We observed that the Ir1-catalyzed intracellular reactions generated TONs of ~5 for 2c and ~6 for 3c, suggesting that poor selectivity was observed. In contrast, the Ir6-catalyzed intracellular reactions afforded TONs of ~1 for 2c and ~14 for 3c, showing a clear preference for the primary amine over alcohol product. The greater selectivity of Ir6 relative to that of Ir1 in RA is consistent with our studies in solution (Figure 2B).

Reductive Amination in Living Cells: Eliciting Inhibition Effects.

Building on the success of our cell experiments above, we sought to demonstrate a scenario in which in-cell drug synthesis would result in cell growth inhibition. Cinacalcet, an FDA-approved drug containing a secondary amine moiety, was selected as our model compound.¹⁷ These studies were conducted in SH-SY5Y cells because there is literature precedent indicating that cinacalcet can reduce the proliferation of neuroblastoma cells, which can be useful in anti-cancer treatment.⁶⁶ Because the RA precursors 1p (IC₅₀ = 175 μM) and (R)-41 (IC₅₀ = 1.5 mM) are less cytotoxic than cinacalcet (IC₅₀ = 16 μM) (Figure S38), we anticipated that cell viability would be reduced upon formation of the drug molecule in situ (Figure 8A). In these studies, we treated SH-SY5Y cells with 1p (200 μM) and HCOONa (1 mM) for 3 h, washed with fresh media, and then added (R)-41 and Ir catalyst for 24 h (Figure 8B and S40). Cells that were exposed to only 1p, 1p/(R)-41, (R)-41/Ir1, or (R)-41/Ir6 were not affected to a substantial degree ( ≥85% cell viability) because one or more reaction components were lacking to enable RA. However, when the aldehyde, nitrogen donor, Ir catalyst, and hydride source were present, the cell viability dropped to 28% in the Ir1-treated group and 46% in the Ir6-treated group relative to that in the untreated control. The wells containing Ir3 or Ir7 as the catalyst for RA reactions also showed reduced cell growth (Figure S40). These results suggest that intracellular Ir-based catalysis has taken place, leading to the cytotoxic accumulation of cinacalcet.

Figure 8.

A) Diagram showing treatment of SH-SY5Y human neuroblastoma cells with aldehyde 1p and amine (R)-41, followed by Ir-catalyzed reduction to the corresponding drug molecule cinacalcet (Cin). B) Cell viability data obtained from carrying out reductive amination of 1p using an Ir catalyst and (R)-41 inside SH-SY5Y cells. C) HPLC analysis of products generated inside cells after treatment with 1p, (R)-41, Ir1 or Ir6, and HCOONa. The turnover numbers (TONs) were calculated by dividing the product concentration by the intracellular iridium concentration. The identity of the peak marked with an asterisk (*) is currently unknown.

To obtain sufficient quantities of products for HPLC analysis, we conducted studies in four 100 mm tissue culture plates by treating SH-SY5Y cells with 1p, (R)-41, Ir1 or Ir6, and HCOONa using our optimized protocol. Upon lysing the cells and analyzing the contents by HPLC, we observed peaks at 3.66 min corresponding to the RA product cinacalcet (Figure 8C). The identity of a minor peak at 3.56 min in the Ir6-treated sample has not yet been determined. Using these data, the TONs for cinacalcet were determined to be ~20 for Ir1 and ~12 for Ir6. Based on our previous studies, the Ir1-H species is more hydridic than Ir6-H,⁷⁸ so the former is expected to reduce the cinacalcet imine intermediate to the 2° amine product faster than the latter. The presence of greater amounts of drug molecules in the Ir1-treated cells relative to that in the Ir6-treated cells is reflected in the cell viability data (Figure 8B). Our results demonstrate unambiguously that Ir-catalyzed intracellular RA can be achieved inside living cells to elicit the desired biological response.

Conclusions

In summary, this work presents the first method for the RA of aldehydes to 1°, 2°, or 3° amines under physiological relevant conditions and within the complex environments of living cells. We overcame the overalkylation issue associated with the preparation of 1° amines by employing 4-(1-aminoethyl)phenol as a self-immolative nitrogen donor, which can spontaneously dissemble following RA with carbonyl-containing substrates. The synthesis of 2° and 3° amines was accomplished with moderate to high selectivity using conventional nitrogen donors, without relying on a self-immolation strategy. Our results suggest that electron-poor Ir catalysts are more effective than their electron-rich counterparts in achieving high yields of amine over alcohol products, allowing the aldehydes to react with the aminating agents faster than their direct reduction by the respective Ir-H species. Furthermore, the aldehyde side chains in deaminated bovine serum albumin could be converted back to the primary amine through Ir-catalyzed RA. To demonstrate the feasibility of intracellular RA, our method was used to produce the alkaloid phenethylamine and calcium-reducing drug cinacalcet in mammalian cells. The in-cell generation of the desired amines was confirmed by the associated biological responses and by HPLC characterization and quantification of the organic products extracted from the cells. Given the importance of the amino group in numerous bioactive molecules, we anticipate that our iridium-catalyzed RA method will find broad utility in a wide range of biological and therapeutic applications. However, further studies are needed to assess the biorthogonality of this approach, including its potential to alkylate endogenous species within living cells. Although it has been reported that Ir-H species can react with oxidized nicotinamide adenine dinucleotide (NAD⁺) to generate the reduced cofactor NADH,^56,120 this transformation has not yet been investigated using our electron-poor catalyst variants. Finally, this research contributes to a growing body of work demonstrating that synthetic reactions developed in the flask can be successfully translated to cellular systems by prioritizing biocompatibility as a reaction design parameter.

Supplementary Material

The Supporting Information is available free of charge on the ACS Publications website.

Synthetic and experimental procedures, reaction yields, kinetic analysis, cell studies, biological assay results (PDF)