Understanding Structural Isomerism in Organoiridium Picolinamidate Complexes and its Consequences on Reactivity and Biological Properties.

Organoiridium picolinamidate complexes are promising for intracellular applications because of their biocompatibility, activity in living systems, and ease of derivatization. To shield their metal centers from inhibition by biological nucleophiles (e.g., glutathione), attempts were made to increase the steric bulk of the supporting N -(2,6-R 2 -phenyl)picolinamidate ligand. It was found that when R = H ( Ir1 ) or methyl ( Ir2 ), the ligand adopts N,N′ -coordination to iridium, whereas when R = isopropyl ( Ir3 ) or phenyl ( Ir4 ), N , O -coordination was observed. Based on experimental measurements and density functional theory calculations, it was revealed that the carbon chemical shift of the C(O)NR group can be used as a diagnostic handle to distinguish between the N,N′ - and N , O -isomers in solution. Computational studies indicate that the former is favored thermodynamically but the latter is preferred when the R group is overly bulky. Complexes Ir1 - Ir4 exhibit differences in lipophilicity, cellular uptake, cytotoxicity, and the propensity to generate reactive oxygen species in living cells. Reaction studies showed that Ir1 / Ir2 are more efficient than Ir3 / Ir4 in promoting the reduction of aldehydes to alcohols via transfer hydrogenation but both isomer types were susceptible to catalyst poisoning by glutathione. This work has led to new insights into structural isomerism in organoiridium picolinamidate complexes and suggests that steric tuning alone is insufficient to protect the Ir center from poisoning by biological nucleophiles. ## Introduction Introduction Half-sandwich metal complexes have been used in numerous biological applications, including as therapeutic agents, 1 – 3 intracellular catalysts, 4 – 9 and cofactors for artificial metalloenzymes. 10 They typically comprise metal arenes (where M = Ru, Rh, Os, or Ir) coordinated by a multidentate ligand and at least one labile X donor (e.g., chloride). In biological media, the M–X complex can be aquated to yield M–OH 2 species or attacked by glutathione (GSH) to generate M–SG species ( Chart 1A ). 11 , 12 Because the metal-sulfur interaction is strong, binding of GSH to M can interfere with its catalytic abilities by occupying a coordination site. 13 , 14 Given that GSH and other endogenous nucleophiles (e.g., nitrogenous bases) are prevalent in living systems, strategies to prevent their coordination to small-molecule metal catalysts would significantly enhance their intracellular activity. To prevent the inhibition of half-sandwich organometallic complexes, a variety of methods have been explored. It was reported that the use of substitutionally inert ancillary ligands can protect metal catalysts against GSH attack. For example, Sadler and coworkers demonstrated that replacing chloride with pyridine in a Cp*Ir complex (Cp* = pentamethylcyclopentadienyl anion, Cat1 ) led to more productive reactive oxygen species (ROS) generation and cancer cell death compared to that of the parent catalyst ( Chart 1B ). 11 , 15 It was proposed that because Cat1 is slow to react with external donors (e.g., H 2 O or GSH), it was able to reach its biological targets more effectively. Expanding on this concept, Pizarro and coworkers designed a series of organoiridium catalysts bearing a pyridine ring tethered to Cp* ( Cat2 , Chart 1B ). 16 – 18 The investigators showed that the pyridine-tethered catalysts were substantially more potent against cancer cells than their non-tethered variants. The enhanced effects were attributed to the tethered pyridine being able to shield the Ir center from external nucleophiles and minimize premature catalyst activation. Finally, Das and Paira reported that a Cp*Ir complex ( Cat3 ) with a coordinated 1,3,5-triaza-7-phosphaadamantane (PTA) was capable of achieving high phototoxicity against triple-negative breast cancer cells ( Chart 1B ). 19 Once again, the slow dissociation of the ancillary ligand from Ir was invoked as the reason for its resistance to GSH inhibition. Although metalloenzymes can restrict access to their active sites using protein channels, 20 – 22 small-molecule metal catalysts lack this ability. However, there has been some success in using bulky ligands to protect the metal centers from reaction with undesired species. For example, Mukherjee and coworkers found that having sterically encumbering 2,6-diisopropyl groups in the ligand framework of the Ru complex Cat4 was key to preventing its deactivation by GSH ( Chart 1C ). 23 Inspired by these studies, we proceeded to explore the application of steric tuning to enhance the GSH tolerance of Cp*Ir picolinamidate complexes ( Cat5 ). 24 – 26 There is growing interest in these catalysts because they are active inside living cells and are amenable to further functionalization for specialized applications. 27 – 29 Attempts to use 2,2′-bipyridine, which is electronically similar to the iminopyridine ligand in Cat4 , with Cp*Ir did not afford active transfer hydrogenation catalysts under mild conditions. 30 , 31 In the present work, we have systematically varied the 2,6-substituents on the N -phenyl ring in Cat5 . Our results showed that increasing the steric bulk in the picolinamidate ligand led to a preference for N , O - rather than N,N′ -coordination to iridium. 32 – 34 The two isomers were distinguishable by NMR spectroscopy and the chemical basis for their existence was studied by density functional theory (DFT) calculations. Although GSH tolerance was not achieved, these investigations offered new insights into the design criteria for constructing Cp*Ir picolinamidate-based systems with high activity. ## Results and Discussion Results and Discussion Synthesis and Characterization of the Ir Complexes. To study the effects of steric hindrance on the reactivity of Cp*Ir picolinamidate complexes, we prepared a series of N -phenylpicolinamides ( 4a - 4f ) with different R substituents at the 2,6-positions of the phenyl ring ( Figure 1A ). The ligands were obtained by combining the desired aniline with 2-picolinic acid using either 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and hydroxybenzotriazole (for R = H ( 4a ), Me ( 4b ), and i Pr ( 4c )) or propylphosphonic anhydride (for R = Ph ( 4e ) or 3,5-Me 2 Ph ( 4f )) as amide coupling reagents ( Scheme S1 ). Attempts to synthesize 4d with R = t- Bu were unsuccessful using either amide coupling reagents. We found that metalation of the picolinamides required different methods depending on the steric bulk of the R group. For example, 4a and 4b were treated with [Cp*IrCl 2 ] 2 and NH 4 PF 6 in EtOH to furnish Ir1 and Ir2 with ≥70% yield, respectively (Method A). The preparation of Ir1 was reported previously. 24 In contrast, the bulkier ligands 4c and 4e were metalated by reaction with [Cp*IrCl 2 ] 2 and NEt 3 in CH 2 Cl 2 to provide Ir3 and Ir4 , respectively (Method B). The yields were modest (≤40%), potentially due to the slower rates of reactions as a result of increased steric hindrance. The bulkiest ligand, 4f , could not be metalated using either Methods A or B ( Scheme S1 ). Next, the iridium complexes were characterized using a variety of spectroscopic methods. We observed that Ir1 - Ir4 displayed strong absorbances below 450 nm, similar to other organoiridium picolinamidate complexes reported in the literature ( Figure 1B , left). 31 They also exhibit multiple vibrational features around 1560–1630 cm −1 , most likely due to the presence of the C(O)NR group ( Figure 1B , right). 35 These spectroscopic data, however, do not provide information about the structure or atom connectivity of the Ir species. To obtain further insights into the nature of the Ir complexes in solution, we used NMR spectroscopy. The 1 H NMR spectra of Ir1 ( Figure S33 ) and Ir4 ( Figure S39 ) showed sharp resonances in the aliphatic and aromatic regions, indicating that the Ir center is coordinated to Cp* and the picolinamidate ligand, respectively. We observed that in the 1 H NMR spectrum of Ir2 , there are two signals centered at 2.1 and 2.3 ppm ( Figure 1C , top), which were assigned to the methyl substituents of the picolinamidate ligand. Because these peaks are broad and magnetically inequivalent, it suggests that the 2,6-dimethylphenyl group has restricted rotational freedom. In contrast, the 1 H NMR spectrum of Ir3 exhibits two sharp signals (~1.1 and 1.2 ppm) corresponding to the -CH 3 of the isopropyl groups ( Figure 1C , bottom). If the 2,6-diisopropylphenyl ring was conformationally rigid, four signals originating from the -CH 3 units would be expected. Thus, these results suggest that the 2,6-diisopropylphenyl moiety in Ir3 has greater rotational freedom than the 2,6-dimethylphenyl moiety in Ir2 , which is counterintuitive based on their relative steric volumes. The surprising observation above led us to compare the 13 C NMR spectra of the iridium complexes more closely. We found that the carbon chemical shifts of the C(O)NR group appeared at 168.4, 169.9, 163.0, and 163.8 ppm in Ir1 , Ir2 , Ir3 , and Ir4 , respectively ( Figure 2 ). A survey of the literature revealed that all reported Cp*Ir complexes with confirmed or presumed N,N′ -coordination of the picolinamidate ligand to Ir exhibit C(O)NR peaks at ≥167 ppm, regardless of electronic differences in their ligands (e.g., complexes I - VI in Figures 2A ; additional examples in Figure S55 ). 24 , 25 , 28 , 31 , 36 – 42 Given that the C(O)NR resonances in Ir3 and Ir4 clearly fall outside of the expected range, we suspected that they are structurally distinct from the N,N′ -coordinated Cp*Ir picolinamidate species. To elucidate the structures of the iridium complexes, we grew single crystals of Ir1 - Ir4 and analyzed them by X-ray crystallography ( Table S9 ). We found that all four complexes have the formula [Cp*Ir(picolinamidate)Cl] but they differ in the binding mode of their supporting ligands. Complexes Ir1 ( Figure S47 ) and Ir2 ( Figure 3A ) feature N,N′ -coordination whereas Ir3 ( Figure 3A ) and Ir4 ( Figure S50 ) feature N , O -coordination to the metal. Based on the bond distances ( Figure 3B ), the C(O)NR group in Ir1 and Ir2 is best described as an amidate with C–O and C–N bond distances of ~1.24 and ~1.33 Å, respectively. 43 In contrast, the C(O)NR group in Ir3 and Ir4 has bond metrics that are consistent with an iminolate group (C–O = 1.30 Å and C–N = 1.29 Å). 43 In light of these results, our NMR data above can be further clarified. Because Ir2 exists as the N,N′ -isomer, its 2,6-dimethylphenyl group has limited rotational freedom due to possible steric clash with the Cp* ring. However, the 2,6-diisopropylphenyl group in Ir3 has free rotation because it is positioned further away from Cp* in the N,O -isomer structure. The percent buried volumes (%V bur ) of the Ir complexes, 44 which measure the steric bulk of the picolinamidate ligands, corroborate the finding that N,O -binding results in reduced steric encumbrance relative to N,N′ -binding ( Figure S46 and Table S8 ) Density Functional Theory Calculations. To verify the binding mode of the structural isomers, we turned to density functional theory (DFT) to compute the NMR chemical shifts ( δ ) for the N,N′ - and N,O - forms of the iridium complexes. The chemical shifts were computed at the SMD-(CH 2 Cl 2 )-B97–2//M11/def2-TZVPP level, 45 and the results for select diagnostic peaks (proton at the 3-position of pyridine and carbon in the C(O)NR unit) are provided in Figure 2B . Our studies revealed that the computed 1 H chemical shifts matched the experimental chemical shifts better for the N,N′ -isomer of Ir1 / Ir2 and the N , O -isomer of Ir3 / Ir4 . For example, the absolute deviation (Δ δ ) was 0.09 and 0.22 ppm for the N,N′ -isomers of Ir1 and Ir2 , respectively, which is smaller than that for their corresponding N,O -isomers. In contrast, the Δ δ for Ir3 and Ir4 was 0.53 ppm for the N,N′ -isomer and ≤0.23 ppm for the N , O -isomer. Given that the tolerable error for the computed δ is <0.3 ppm, the N , O -isomer was clearly a closer match than the N,N′ -isomer for Ir3 and Ir4 . Comparison of the computed vs. experimentally measured carbon chemical shifts showed similar trends ( Figure 2B ). We found that the Δ δ for the diagnostic carbon in the N,N′ -isomer of Ir1 and Ir2 was ≤0.4 ppm whereas that in the corresponding N , O -isomer was ≥2.0 ppm. For Ir3 and Ir4 , the opposite was observed, in which the Δ δ for the N,N′ -isomer was larger (≥7.7 ppm) than that for the N , O -isomer (≤1.9 ppm). These results suggest that Ir1 and Ir2 exists exclusively in the N,N′ -form whereas Ir3 and Ir4 exist exclusively in the N,O -form, which is consistent with our crystallographic data. Studies of Ir3 using variable temperature NMR spectroscopy in CD 3 OD did not show any changes from −35 to 25 °C, indicating that it cannot interconvert between different structural isomers in solution under the conditions tested. It should also be noted that the upfield shifted carbon peaks assigned to the C(O)NR group in Ir3 and Ir4 , relative to that in Ir1 and Ir2 , were reproduced computationally at 164.9 (Δ δ = 1.9) and 165.6 ppm (Δ δ =1.8), respectively. Thus, it appears that the carbon chemical shift of the C(O)NR unit can be used as a diagnostic handle to differentiate between N,N′ - and N,O -coordination in these systems ( Figure 2A ). To understand the chemical basis for the N,N′ - vs. N , O - binding preference, the relative Gibbs free energies of the iridium complexes were determined by DFT. In all cases, the N,N′ -isomers were energetically more favorable than the N , O -isomers by 8.9, 6.7, 5.4, and 1.7 kcal/mol for Ir1 , Ir2 , Ir3 , and Ir4 , respectively ( Figure 4A ). Although these results would suggest that only the N,N′ -isomer should exist, further DFT studies revealed that steric hindrance may be responsible for the isomer preference. Our results showed that the N,N′ -isomer of Ir1 has minimum repulsive noncovalent interactions (NCIs) whereas the N,N′ -isomers of Ir3 and Ir4 have several regions of significant steric clash ( Figures 4B and S51 – S54 ). 46 Specifically, the R groups (i.e., i Pr in Ir3 and Ph in Ir4 ) have unfavorable NCIs with the Cp* ring and the amidate moiety in the case of Ir4 . In contrast, the N,O -isomers of Ir3 and Ir4 exhibit mostly attractive and van der Waals forces ( Figures S53 – S54 ). Thus, our results revealed that although the N,N′ -isomer is more thermodynamically stable than the N,O -isomer, the presence of bulky R groups leads to a preference for the latter to minimize steric repulsion. Transfer Hydrogenation Studies. Our new iridium complexes offered an opportunity to compare the reactivity of the N,N′ - vs. N , O -isomers. Using our standard conditions for transfer hydrogenation reported in previous work, 30 , 31 we evaluated the ability of Ir1 - Ir4 to catalyze the conversion of aldehydes to alcohols using HCOONa in DMSO/H 2 O (1:9) under air at 37 °C ( Table 1 ). After 24 h, Ir1 and Ir2 gave quantitative yields of benzyl alcohol starting from benzaldehyde but Ir3 produced only 32% yield and Ir4 was inactive (entry 1). A time-dependence study using Ir1 and Ir2 showed that the former is more active than the latter ( Table S2 ), suggesting that when R = Me, the increased steric bulk may lead to slower catalysis. The lower activity of the N,O -isomers ( Ir3 / Ir4 ) compared to the N,N′- isomers ( Ir1 / Ir2 ) is likely due to their more electron poor nature, which has been shown to reduce the efficiency of hydride transfer from the Ir–H intermediate to aldehyde acceptors. 31 In the case of Ir4 , we found that it does not react readily with HCOONa, which is the first step in the transfer hydrogenation process. We hypothesize that the excessive steric bulk in Ir4 due to the presence of the phenyl substituents is most likely responsible for its inefficient reactions with both formate and substrates. We observed that the efficiency of benzaldehyde reduction using the Ir catalysts and HCOONa is dependent on the reaction medium. For example, reactions in DMSO/Dulbecco’s Modified Eagle Medium (DMEM) (1:9) gave similar results as those in DMSO/H 2 O (1:9) ( Table S1 ). However, the addition of 10% fetal bovine serum (FBS) to the DMSO/DMEM mixture afforded lower yields of benzyl alcohol (90% for Ir1 , 95% for Ir2 , 22% for Ir3 , and 0% for Ir4 ; Table S1 ). The use of DMSO/Roswell Park Memorial Institute-1640 (RPMI-1640) also lowered the transfer hydrogenation efficiency. Consistent with our previous findings, 47 the presence of biological components and nucleophiles can coordinatively inhibit the iridium catalysts. When the reactions were performed in the presence of 1.0 mM GSH ( Table 1 , entry 2), the yields of benzyl alcohol for Ir1 and Ir2 dropped substantially compared to that in the absence of GSH and Ir3 was completely inhibited. Although Ir2 was slightly more tolerant of GSH than Ir1 (27 vs. 18% yield, respectively), these results suggest that using bulky ligands to protect the Ir centers in Cp*Ir picolinamidate catalysts from coordination inhibition is challenging because the less active N,O -isomer is generated if the R group is too bulky. The iridium complexes were next tested for their transfer hydrogenation activity toward biologically relevant aldehydes. 48 Under our standard conditions using HCOONa as the hydride source, Ir1 , Ir2 , and Ir3 fully reduced crotonaldehyde to 1-butanol with ≥95% yield ( Table 1 , entry 3). In contrast, Ir4 afforded the partially reduced species butanal and but-2-en-1-ol in 9% and 7% yield, respectively ( Table S3 ). When hexanal was employed as the substrate, Ir1 , Ir2 , and Ir3 gave quantitative amounts of hexan-1-ol whereas Ir4 gave only a 9% yield ( Table 1 , entry 4). In general, these studies demonstrated that the N,N′ -isomers are more active than the N,O -isomers for transfer hydrogenation and that having sterically-encumbering R groups in the picolinamidate ligand decrease the catalytic efficiency. We noticed that in the reactions above, the amount of HCOONa consumed was consistently greater than the amount of substrate reduced. Because it has been shown in previous studies that Cp*Ir catalysts are capable of converting O 2 to H 2 O 2 in the presence of HCOONa, 11 , 28 we measured the hydrogen peroxide levels under various reaction conditions using semi-quantitative color test strips. When the Ir complexes and excess HCOONa were stirred in DMSO/H 2 O (1:9) under air at 37 °C, formation of H 2 O 2 was detected in all reactions ( Figure 5 , left). Complex Ir2 generated significantly more H 2 O 2 than the other Ir catalysts (up to ~300 μM) and the H 2 O 2 concentration fluctuated over the course of 72 h. The changing amounts of H 2 O 2 over time may be due to its disproportionation. 49 , 50 These results indicate that formate dehydrogenation to generate Ir-H species can occur in all four Ir catalysts, albeit with different rates, but the subsequent hydride transfer from Ir-H to O 2 is more facile than to aldehyde substrates. When the reactions were repeated in the presence of benzaldehyde, similar results were obtained ( Figure 5 , right). The most noticeable difference was that the maximum concentration of H 2 O 2 generated by Ir2 was lowered to ~200 μM, which is likely due to competitive reaction of the Ir-H intermediate with both O 2 and the organic electrophile. 28 , 51 , 52 The higher activity of Ir2 relative to that of Ir1 for H 2 O 2 production is the opposite of what was observed for the reduction of benzaldehyde ( Table S2 ). These results can be rationalized by the fact that Ir1 is a less bulky catalyst than Ir2 so it can react more readily with larger substrates, leading to greater reaction efficiency with small-molecule aldehydes (and vice versa). These studies suggest that the iridium catalysts may be capable of producing ROS inside living cells since endogenous reduced nicotinamide adenine dinucleotide (NADH) can serve as a hydride source. 11 , 53 Biological Properties. Because we are interested in using the iridium complexes for applications in living systems, 4 , 5 , 7 , 54 we proceeded to evaluate their reactions with biologically-relevant species. To probe their tendency to be aquated, the Ir complexes were dissolved in DMSO- d 6 and then their 1 H NMR spectra were recorded before and after the addition of water. We observed that in the presence of up to 0.5 equiv. of H 2 O, relative to Ir, no significant changes in the chemical shifts of Ir2 had occurred ( Figure S41 ), indicating that it remained in the Ir–Cl form. In contrast, when Ir4 was combined with 0.5 equiv. of H 2 O, two distinct sets of peaks appeared in the NMR spectra ( Figures S42 and S43 ). The Cp* peak at 1.4 ppm was assigned to the starting Ir–Cl species whereas the Cp* peak at 1.2 ppm was assigned to the newly formed Ir–OH 2 species. These results indicate that N,N′ - and N,O -isomers have different propensities toward aquation. Next, the interactions of Ir1 , Ir2 , and Ir3 with endogenous nucleophiles in DMSO/H 2 O (1:9) were investigated by UV-vis absorption spectroscopy. We found that the addition of up to 10 equiv. of 2-acetamido-6-hydroxypurine ( Figures S4 – S6 ), cysteine ( Figures S7 – S9 ), or glutathione ( Figures S10 – S12 ), relative to Ir, led to distinct spectral changes. Although the nature of these interactions was not investigated further, coordination of biological nucleophiles to Cp*Ir complexes is well documented. 13 , 14 To estimate the lipophilicity of the iridium complexes, their octanol/water partition coefficients (Log P ) were measured. 55 Sodium chloride (0.25 M) was added to the aqueous solutions to mimic the high salt concentrations of biological media. Our data revealed that the Ir complexes had Log P values in the order Ir4 (1.32)> Ir3 (0.88)> Ir2 (0.78)> Ir1 (0.48) ( Table 2 , column 2). As expected, when the R group of the picolinamidate ligand is alkyl or Ph, the Ir complexes’ overall hydrophobicity increases relative to that of the parent complex Ir1 . Next, the cellular uptake of the iridium complexes in NIH-3T3 mouse fibroblast cells was determined. This cell line was chosen due to its robustness and common use in mammalian cell studies. 56 – 58 The cells were incubated with 10 μM of Ir1 , Ir2 , Ir3 , or Ir4 for 24 h, washed with fresh media, and then lysed so that the iridium content of the cells could be analyzed by inductively coupled plasma mass spectrometry (ICP-MS). According to our data, the following cell uptake efficiency was observed: Ir4 > Ir2 > Ir1 ~ Ir3 , with intracellular Ir concentrations ranging from 45–163 ng/10 6 cells ( Table 2 , column 3). Although the cell uptake trend does not correlate with the Log P values, differences in the speciation or cell uptake mechanism of the iridium complexes may account for this discrepancy. Given that the iridium complexes are readily internalized in the NIH-3T3 cells, we next evaluated their cytotoxicity. For these experiments, the cells were exposed to various concentrations of the iridium complexes for 24 h and then their cell viability was measured using a sulforhodamine B (SRB) assay. The 50% inhibition concentrations (IC 50 ) were found to be 71, 45, >200, and 60 μM for Ir1 , Ir2 , Ir3 , and Ir4 , respectively ( Table 2 , column 4). The cytotoxicity of the iridium complexes does not appear to correlate with either their Log P values or cell uptake efficiency. To assess whether the cytotoxicity of the Ir catalysts is cell line dependent, we also screened them using A549 human lung cancer cells and BEAS-2B human lung non-cancer cells ( Table S15 ). Our results showed that in A549, the IC 50 values ranged from ~45 to 129 μM in the order Ir2 < Ir4 < Ir1 < Ir3 , which is consistent with the cytotoxicity trend observed in NIH-3T3 cells. However, in the BEAS-2B cell line, the IC 50 values were observed in the order Ir4 < Ir1 < Ir3 ~ Ir2 ( Table S15 ). For comparison, the IC 50 of Cat1 was reported to be ~4 μM in A549 cells. 15 Because there are many factors that contribute to cytotoxicity, further studies are needed to understand the origin of these different effects. Although the N,N′- and N,O -isomers exhibit clear differences in their reactivity, as demonstrated in the transfer hydrogenation section above, their biological properties are not dependent on their picolinamidate ligand binding mode. Thus, the characteristics of each iridium complex inside cells must be measured individually rather than relying on their molecular structures to predict their biological behavior. Finally, we wanted to assess whether the ROS-generating ability of the Ir complexes could be predicted based on their reactivity in solution. 11 In these studies, NIH-3T3 cells were incubated with 5 μM of an iridium complex and 2 mM of HCOONa (if any) for 24 h prior to measuring the ROS levels using a fluorescein-based assay ( Figure S17 ). We observed that cells containing Ir1 , Ir2 , and Ir3, with or without HCOONa, displayed higher amounts of ROS compared to those in the untreated control. No statistically significant increase in ROS was found in cells exposed to Ir4 . It should be noted that Ir2 produced the largest amounts of ROS in cells ( Table 2 , column 5), which is consistent with our studies showing that it was also the most active for H 2 O 2 production in the reaction flask ( Figure 5 ). Thus, the reactivity of the iridium complexes with O 2 in solution appears to roughly mirror their relative reaction rates in cells. However, this observation is likely true only under limited circumstances since the efficiency of intracellular reactions is dependent on numerous factors (e.g., local pH, GSH concentrations, catalyst localization, etc.). 7 ## Synthesis and Characterization of the Ir Complexes. Synthesis and Characterization of the Ir Complexes. To study the effects of steric hindrance on the reactivity of Cp*Ir picolinamidate complexes, we prepared a series of N -phenylpicolinamides ( 4a - 4f ) with different R substituents at the 2,6-positions of the phenyl ring ( Figure 1A ). The ligands were obtained by combining the desired aniline with 2-picolinic acid using either 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride and hydroxybenzotriazole (for R = H ( 4a ), Me ( 4b ), and i Pr ( 4c )) or propylphosphonic anhydride (for R = Ph ( 4e ) or 3,5-Me 2 Ph ( 4f )) as amide coupling reagents ( Scheme S1 ). Attempts to synthesize 4d with R = t- Bu were unsuccessful using either amide coupling reagents. We found that metalation of the picolinamides required different methods depending on the steric bulk of the R group. For example, 4a and 4b were treated with [Cp*IrCl 2 ] 2 and NH 4 PF 6 in EtOH to furnish Ir1 and Ir2 with ≥70% yield, respectively (Method A). The preparation of Ir1 was reported previously. 24 In contrast, the bulkier ligands 4c and 4e were metalated by reaction with [Cp*IrCl 2 ] 2 and NEt 3 in CH 2 Cl 2 to provide Ir3 and Ir4 , respectively (Method B). The yields were modest (≤40%), potentially due to the slower rates of reactions as a result of increased steric hindrance. The bulkiest ligand, 4f , could not be metalated using either Methods A or B ( Scheme S1 ). Next, the iridium complexes were characterized using a variety of spectroscopic methods. We observed that Ir1 - Ir4 displayed strong absorbances below 450 nm, similar to other organoiridium picolinamidate complexes reported in the literature ( Figure 1B , left). 31 They also exhibit multiple vibrational features around 1560–1630 cm −1 , most likely due to the presence of the C(O)NR group ( Figure 1B , right). 35 These spectroscopic data, however, do not provide information about the structure or atom connectivity of the Ir species. To obtain further insights into the nature of the Ir complexes in solution, we used NMR spectroscopy. The 1 H NMR spectra of Ir1 ( Figure S33 ) and Ir4 ( Figure S39 ) showed sharp resonances in the aliphatic and aromatic regions, indicating that the Ir center is coordinated to Cp* and the picolinamidate ligand, respectively. We observed that in the 1 H NMR spectrum of Ir2 , there are two signals centered at 2.1 and 2.3 ppm ( Figure 1C , top), which were assigned to the methyl substituents of the picolinamidate ligand. Because these peaks are broad and magnetically inequivalent, it suggests that the 2,6-dimethylphenyl group has restricted rotational freedom. In contrast, the 1 H NMR spectrum of Ir3 exhibits two sharp signals (~1.1 and 1.2 ppm) corresponding to the -CH 3 of the isopropyl groups ( Figure 1C , bottom). If the 2,6-diisopropylphenyl ring was conformationally rigid, four signals originating from the -CH 3 units would be expected. Thus, these results suggest that the 2,6-diisopropylphenyl moiety in Ir3 has greater rotational freedom than the 2,6-dimethylphenyl moiety in Ir2 , which is counterintuitive based on their relative steric volumes. The surprising observation above led us to compare the 13 C NMR spectra of the iridium complexes more closely. We found that the carbon chemical shifts of the C(O)NR group appeared at 168.4, 169.9, 163.0, and 163.8 ppm in Ir1 , Ir2 , Ir3 , and Ir4 , respectively ( Figure 2 ). A survey of the literature revealed that all reported Cp*Ir complexes with confirmed or presumed N,N′ -coordination of the picolinamidate ligand to Ir exhibit C(O)NR peaks at ≥167 ppm, regardless of electronic differences in their ligands (e.g., complexes I - VI in Figures 2A ; additional examples in Figure S55 ). 24 , 25 , 28 , 31 , 36 – 42 Given that the C(O)NR resonances in Ir3 and Ir4 clearly fall outside of the expected range, we suspected that they are structurally distinct from the N,N′ -coordinated Cp*Ir picolinamidate species. To elucidate the structures of the iridium complexes, we grew single crystals of Ir1 - Ir4 and analyzed them by X-ray crystallography ( Table S9 ). We found that all four complexes have the formula [Cp*Ir(picolinamidate)Cl] but they differ in the binding mode of their supporting ligands. Complexes Ir1 ( Figure S47 ) and Ir2 ( Figure 3A ) feature N,N′ -coordination whereas Ir3 ( Figure 3A ) and Ir4 ( Figure S50 ) feature N , O -coordination to the metal. Based on the bond distances ( Figure 3B ), the C(O)NR group in Ir1 and Ir2 is best described as an amidate with C–O and C–N bond distances of ~1.24 and ~1.33 Å, respectively. 43 In contrast, the C(O)NR group in Ir3 and Ir4 has bond metrics that are consistent with an iminolate group (C–O = 1.30 Å and C–N = 1.29 Å). 43 In light of these results, our NMR data above can be further clarified. Because Ir2 exists as the N,N′ -isomer, its 2,6-dimethylphenyl group has limited rotational freedom due to possible steric clash with the Cp* ring. However, the 2,6-diisopropylphenyl group in Ir3 has free rotation because it is positioned further away from Cp* in the N,O -isomer structure. The percent buried volumes (%V bur ) of the Ir complexes, 44 which measure the steric bulk of the picolinamidate ligands, corroborate the finding that N,O -binding results in reduced steric encumbrance relative to N,N′ -binding ( Figure S46 and Table S8 ) ## Density Functional Theory Calculations. Density Functional Theory Calculations. To verify the binding mode of the structural isomers, we turned to density functional theory (DFT) to compute the NMR chemical shifts ( δ ) for the N,N′ - and N,O - forms of the iridium complexes. The chemical shifts were computed at the SMD-(CH 2 Cl 2 )-B97–2//M11/def2-TZVPP level, 45 and the results for select diagnostic peaks (proton at the 3-position of pyridine and carbon in the C(O)NR unit) are provided in Figure 2B . Our studies revealed that the computed 1 H chemical shifts matched the experimental chemical shifts better for the N,N′ -isomer of Ir1 / Ir2 and the N , O -isomer of Ir3 / Ir4 . For example, the absolute deviation (Δ δ ) was 0.09 and 0.22 ppm for the N,N′ -isomers of Ir1 and Ir2 , respectively, which is smaller than that for their corresponding N,O -isomers. In contrast, the Δ δ for Ir3 and Ir4 was 0.53 ppm for the N,N′ -isomer and ≤0.23 ppm for the N , O -isomer. Given that the tolerable error for the computed δ is <0.3 ppm, the N , O -isomer was clearly a closer match than the N,N′ -isomer for Ir3 and Ir4 . Comparison of the computed vs. experimentally measured carbon chemical shifts showed similar trends ( Figure 2B ). We found that the Δ δ for the diagnostic carbon in the N,N′ -isomer of Ir1 and Ir2 was ≤0.4 ppm whereas that in the corresponding N , O -isomer was ≥2.0 ppm. For Ir3 and Ir4 , the opposite was observed, in which the Δ δ for the N,N′ -isomer was larger (≥7.7 ppm) than that for the N , O -isomer (≤1.9 ppm). These results suggest that Ir1 and Ir2 exists exclusively in the N,N′ -form whereas Ir3 and Ir4 exist exclusively in the N,O -form, which is consistent with our crystallographic data. Studies of Ir3 using variable temperature NMR spectroscopy in CD 3 OD did not show any changes from −35 to 25 °C, indicating that it cannot interconvert between different structural isomers in solution under the conditions tested. It should also be noted that the upfield shifted carbon peaks assigned to the C(O)NR group in Ir3 and Ir4 , relative to that in Ir1 and Ir2 , were reproduced computationally at 164.9 (Δ δ = 1.9) and 165.6 ppm (Δ δ =1.8), respectively. Thus, it appears that the carbon chemical shift of the C(O)NR unit can be used as a diagnostic handle to differentiate between N,N′ - and N,O -coordination in these systems ( Figure 2A ). To understand the chemical basis for the N,N′ - vs. N , O - binding preference, the relative Gibbs free energies of the iridium complexes were determined by DFT. In all cases, the N,N′ -isomers were energetically more favorable than the N , O -isomers by 8.9, 6.7, 5.4, and 1.7 kcal/mol for Ir1 , Ir2 , Ir3 , and Ir4 , respectively ( Figure 4A ). Although these results would suggest that only the N,N′ -isomer should exist, further DFT studies revealed that steric hindrance may be responsible for the isomer preference. Our results showed that the N,N′ -isomer of Ir1 has minimum repulsive noncovalent interactions (NCIs) whereas the N,N′ -isomers of Ir3 and Ir4 have several regions of significant steric clash ( Figures 4B and S51 – S54 ). 46 Specifically, the R groups (i.e., i Pr in Ir3 and Ph in Ir4 ) have unfavorable NCIs with the Cp* ring and the amidate moiety in the case of Ir4 . In contrast, the N,O -isomers of Ir3 and Ir4 exhibit mostly attractive and van der Waals forces ( Figures S53 – S54 ). Thus, our results revealed that although the N,N′ -isomer is more thermodynamically stable than the N,O -isomer, the presence of bulky R groups leads to a preference for the latter to minimize steric repulsion. ## Transfer Hydrogenation Studies. Transfer Hydrogenation Studies. Our new iridium complexes offered an opportunity to compare the reactivity of the N,N′ - vs. N , O -isomers. Using our standard conditions for transfer hydrogenation reported in previous work, 30 , 31 we evaluated the ability of Ir1 - Ir4 to catalyze the conversion of aldehydes to alcohols using HCOONa in DMSO/H 2 O (1:9) under air at 37 °C ( Table 1 ). After 24 h, Ir1 and Ir2 gave quantitative yields of benzyl alcohol starting from benzaldehyde but Ir3 produced only 32% yield and Ir4 was inactive (entry 1). A time-dependence study using Ir1 and Ir2 showed that the former is more active than the latter ( Table S2 ), suggesting that when R = Me, the increased steric bulk may lead to slower catalysis. The lower activity of the N,O -isomers ( Ir3 / Ir4 ) compared to the N,N′- isomers ( Ir1 / Ir2 ) is likely due to their more electron poor nature, which has been shown to reduce the efficiency of hydride transfer from the Ir–H intermediate to aldehyde acceptors. 31 In the case of Ir4 , we found that it does not react readily with HCOONa, which is the first step in the transfer hydrogenation process. We hypothesize that the excessive steric bulk in Ir4 due to the presence of the phenyl substituents is most likely responsible for its inefficient reactions with both formate and substrates. We observed that the efficiency of benzaldehyde reduction using the Ir catalysts and HCOONa is dependent on the reaction medium. For example, reactions in DMSO/Dulbecco’s Modified Eagle Medium (DMEM) (1:9) gave similar results as those in DMSO/H 2 O (1:9) ( Table S1 ). However, the addition of 10% fetal bovine serum (FBS) to the DMSO/DMEM mixture afforded lower yields of benzyl alcohol (90% for Ir1 , 95% for Ir2 , 22% for Ir3 , and 0% for Ir4 ; Table S1 ). The use of DMSO/Roswell Park Memorial Institute-1640 (RPMI-1640) also lowered the transfer hydrogenation efficiency. Consistent with our previous findings, 47 the presence of biological components and nucleophiles can coordinatively inhibit the iridium catalysts. When the reactions were performed in the presence of 1.0 mM GSH ( Table 1 , entry 2), the yields of benzyl alcohol for Ir1 and Ir2 dropped substantially compared to that in the absence of GSH and Ir3 was completely inhibited. Although Ir2 was slightly more tolerant of GSH than Ir1 (27 vs. 18% yield, respectively), these results suggest that using bulky ligands to protect the Ir centers in Cp*Ir picolinamidate catalysts from coordination inhibition is challenging because the less active N,O -isomer is generated if the R group is too bulky. The iridium complexes were next tested for their transfer hydrogenation activity toward biologically relevant aldehydes. 48 Under our standard conditions using HCOONa as the hydride source, Ir1 , Ir2 , and Ir3 fully reduced crotonaldehyde to 1-butanol with ≥95% yield ( Table 1 , entry 3). In contrast, Ir4 afforded the partially reduced species butanal and but-2-en-1-ol in 9% and 7% yield, respectively ( Table S3 ). When hexanal was employed as the substrate, Ir1 , Ir2 , and Ir3 gave quantitative amounts of hexan-1-ol whereas Ir4 gave only a 9% yield ( Table 1 , entry 4). In general, these studies demonstrated that the N,N′ -isomers are more active than the N,O -isomers for transfer hydrogenation and that having sterically-encumbering R groups in the picolinamidate ligand decrease the catalytic efficiency. We noticed that in the reactions above, the amount of HCOONa consumed was consistently greater than the amount of substrate reduced. Because it has been shown in previous studies that Cp*Ir catalysts are capable of converting O 2 to H 2 O 2 in the presence of HCOONa, 11 , 28 we measured the hydrogen peroxide levels under various reaction conditions using semi-quantitative color test strips. When the Ir complexes and excess HCOONa were stirred in DMSO/H 2 O (1:9) under air at 37 °C, formation of H 2 O 2 was detected in all reactions ( Figure 5 , left). Complex Ir2 generated significantly more H 2 O 2 than the other Ir catalysts (up to ~300 μM) and the H 2 O 2 concentration fluctuated over the course of 72 h. The changing amounts of H 2 O 2 over time may be due to its disproportionation. 49 , 50 These results indicate that formate dehydrogenation to generate Ir-H species can occur in all four Ir catalysts, albeit with different rates, but the subsequent hydride transfer from Ir-H to O 2 is more facile than to aldehyde substrates. When the reactions were repeated in the presence of benzaldehyde, similar results were obtained ( Figure 5 , right). The most noticeable difference was that the maximum concentration of H 2 O 2 generated by Ir2 was lowered to ~200 μM, which is likely due to competitive reaction of the Ir-H intermediate with both O 2 and the organic electrophile. 28 , 51 , 52 The higher activity of Ir2 relative to that of Ir1 for H 2 O 2 production is the opposite of what was observed for the reduction of benzaldehyde ( Table S2 ). These results can be rationalized by the fact that Ir1 is a less bulky catalyst than Ir2 so it can react more readily with larger substrates, leading to greater reaction efficiency with small-molecule aldehydes (and vice versa). These studies suggest that the iridium catalysts may be capable of producing ROS inside living cells since endogenous reduced nicotinamide adenine dinucleotide (NADH) can serve as a hydride source. 11 , 53 ## Biological Properties. Biological Properties. Because we are interested in using the iridium complexes for applications in living systems, 4 , 5 , 7 , 54 we proceeded to evaluate their reactions with biologically-relevant species. To probe their tendency to be aquated, the Ir complexes were dissolved in DMSO- d 6 and then their 1 H NMR spectra were recorded before and after the addition of water. We observed that in the presence of up to 0.5 equiv. of H 2 O, relative to Ir, no significant changes in the chemical shifts of Ir2 had occurred ( Figure S41 ), indicating that it remained in the Ir–Cl form. In contrast, when Ir4 was combined with 0.5 equiv. of H 2 O, two distinct sets of peaks appeared in the NMR spectra ( Figures S42 and S43 ). The Cp* peak at 1.4 ppm was assigned to the starting Ir–Cl species whereas the Cp* peak at 1.2 ppm was assigned to the newly formed Ir–OH 2 species. These results indicate that N,N′ - and N,O -isomers have different propensities toward aquation. Next, the interactions of Ir1 , Ir2 , and Ir3 with endogenous nucleophiles in DMSO/H 2 O (1:9) were investigated by UV-vis absorption spectroscopy. We found that the addition of up to 10 equiv. of 2-acetamido-6-hydroxypurine ( Figures S4 – S6 ), cysteine ( Figures S7 – S9 ), or glutathione ( Figures S10 – S12 ), relative to Ir, led to distinct spectral changes. Although the nature of these interactions was not investigated further, coordination of biological nucleophiles to Cp*Ir complexes is well documented. 13 , 14 To estimate the lipophilicity of the iridium complexes, their octanol/water partition coefficients (Log P ) were measured. 55 Sodium chloride (0.25 M) was added to the aqueous solutions to mimic the high salt concentrations of biological media. Our data revealed that the Ir complexes had Log P values in the order Ir4 (1.32)> Ir3 (0.88)> Ir2 (0.78)> Ir1 (0.48) ( Table 2 , column 2). As expected, when the R group of the picolinamidate ligand is alkyl or Ph, the Ir complexes’ overall hydrophobicity increases relative to that of the parent complex Ir1 . Next, the cellular uptake of the iridium complexes in NIH-3T3 mouse fibroblast cells was determined. This cell line was chosen due to its robustness and common use in mammalian cell studies. 56 – 58 The cells were incubated with 10 μM of Ir1 , Ir2 , Ir3 , or Ir4 for 24 h, washed with fresh media, and then lysed so that the iridium content of the cells could be analyzed by inductively coupled plasma mass spectrometry (ICP-MS). According to our data, the following cell uptake efficiency was observed: Ir4 > Ir2 > Ir1 ~ Ir3 , with intracellular Ir concentrations ranging from 45–163 ng/10 6 cells ( Table 2 , column 3). Although the cell uptake trend does not correlate with the Log P values, differences in the speciation or cell uptake mechanism of the iridium complexes may account for this discrepancy. Given that the iridium complexes are readily internalized in the NIH-3T3 cells, we next evaluated their cytotoxicity. For these experiments, the cells were exposed to various concentrations of the iridium complexes for 24 h and then their cell viability was measured using a sulforhodamine B (SRB) assay. The 50% inhibition concentrations (IC 50 ) were found to be 71, 45, >200, and 60 μM for Ir1 , Ir2 , Ir3 , and Ir4 , respectively ( Table 2 , column 4). The cytotoxicity of the iridium complexes does not appear to correlate with either their Log P values or cell uptake efficiency. To assess whether the cytotoxicity of the Ir catalysts is cell line dependent, we also screened them using A549 human lung cancer cells and BEAS-2B human lung non-cancer cells ( Table S15 ). Our results showed that in A549, the IC 50 values ranged from ~45 to 129 μM in the order Ir2 < Ir4 < Ir1 < Ir3 , which is consistent with the cytotoxicity trend observed in NIH-3T3 cells. However, in the BEAS-2B cell line, the IC 50 values were observed in the order Ir4 < Ir1 < Ir3 ~ Ir2 ( Table S15 ). For comparison, the IC 50 of Cat1 was reported to be ~4 μM in A549 cells. 15 Because there are many factors that contribute to cytotoxicity, further studies are needed to understand the origin of these different effects. Although the N,N′- and N,O -isomers exhibit clear differences in their reactivity, as demonstrated in the transfer hydrogenation section above, their biological properties are not dependent on their picolinamidate ligand binding mode. Thus, the characteristics of each iridium complex inside cells must be measured individually rather than relying on their molecular structures to predict their biological behavior. Finally, we wanted to assess whether the ROS-generating ability of the Ir complexes could be predicted based on their reactivity in solution. 11 In these studies, NIH-3T3 cells were incubated with 5 μM of an iridium complex and 2 mM of HCOONa (if any) for 24 h prior to measuring the ROS levels using a fluorescein-based assay ( Figure S17 ). We observed that cells containing Ir1 , Ir2 , and Ir3, with or without HCOONa, displayed higher amounts of ROS compared to those in the untreated control. No statistically significant increase in ROS was found in cells exposed to Ir4 . It should be noted that Ir2 produced the largest amounts of ROS in cells ( Table 2 , column 5), which is consistent with our studies showing that it was also the most active for H 2 O 2 production in the reaction flask ( Figure 5 ). Thus, the reactivity of the iridium complexes with O 2 in solution appears to roughly mirror their relative reaction rates in cells. However, this observation is likely true only under limited circumstances since the efficiency of intracellular reactions is dependent on numerous factors (e.g., local pH, GSH concentrations, catalyst localization, etc.). 7 ## Conclusions Conclusions We have prepared and characterized a series of Cp*Ir picolinamidate complexes bearing different R groups at the 2,6-positions of the N -phenyl moiety. We observed that the picolinamidate ligand binds to iridium via N,N′ -coordination when R = H or Me but via N,O -coordination when R = i Pr or Ph. The carbon chemical shift of the C(O)NR group is a useful handle for differentiating between the structural isomers in solution, in which peaks at ≥ 167 ppm are typical for N,N′ -isomers and peaks at ~164 ppm are typical for N,O -isomers. DFT studies suggest that the N,N′- isomer is energetically more favorable than the N,O -isomer but the presence of bulky R groups in the picolinamidate ligand leads to a preference for the latter. In terms of their transfer hydrogenation activity for the reduction of benzaldehyde to benzyl alcohol, Ir1 and Ir2 ( N,N′ -isomers) exhibited higher activity than Ir3 and Ir4 ( N,O -isomers). Unfortunately, all the iridium catalysts showed reduced or no activity when GSH was added. The biological properties of the iridium complexes, such as lipophilicity, cell uptake, and cytotoxicity, are not dependent on their picolinamidate ligand binding mode. Thus, it is necessary to evaluate each complex individually to assess their behavior in living systems. Complex Ir2 was found to produce the highest amounts of ROS in cells, which is consistent with our solution studies showing that it was the most active in generating H 2 O 2 from reaction with HCOONa and O 2 in air. In conclusion, this work has led to a better understanding of structural isomerism in Cp*Ir picolinamidate complexes and offers additional design rules for creating new variants. These studies also suggest that steric tuning alone is insufficient to protect the Ir center from poisoning by biological nucleophiles. Thus, additional strategies are needed to prevent the deactivation of small-molecule organometallic catalysts and maximize their catalytic potential inside living systems. ## Supplementary Material Supplementary Material Supplementary Information CIF