Solution Structures of the Prototypical 18 kDa Translocator Protein Ligand, PK 11195, Elucidated with ¹H/¹³C NMR Spectroscopy and Quantum Chemistry

Abstract

Eighteen kilodalton translocator protein (TSPO) is an important target for drug discovery and for clinical molecular imaging of brain and peripheral inflammatory processes. PK 11195 [1a; 1-(2-chlorophenyl)-N-methyl-(1-methylpropyl)-3-isoquinoline carboxamide] is the major prototypical high-affinity ligand for TSPO. Elucidation of the solution structure of 1a is of interest for understanding small-molecule ligand interactions with the lipophilic binding site of TSPO. Dynamic ¹H/¹³C NMR spectroscopy of 1a revealed four quite stable but interconverting rotamers, due to amide bond and 2-chlorophenyl group rotation. These rotamers have been neglected in previous descriptions of the structure of 1a and of the binding of 1a to TSPO. Here, we used quantum chemistry at the level of B3LYP/6-311+G(2d,p) to calculate ¹³C and ¹H chemical shifts for the rotamers of 1a and for the very weak TSPO ligand, N-desmethyl-PK 11195 (1b). These data, plus experimental NMR data, were then used to characterize the structures of rotamers of 1a and 1b in organic solution. Energy barriers for both the amide bond and 2′-chlorophenyl group rotation of 1a were determined from dynamic ¹H NMR to be similar (ca.17 to 18 kcal/mol), and they compared well with those calculated at the level of B3LYP/6-31G*. Furthermore, the computed barrier for Z to E rotation is considerably lower in 1a(18.7 kcal/mol) than in 1b (25.4 kcal/mol). NMR (NOE) unequivocally demonstrated that the E rotamer of 1a is the more stable in solution by about 0.4 kcal/mol. These detailed structural findings will aid future TSPO ligand design and support the notion that TSPO prefers to bind ligands as amide E-rotamers.

PK 11195 (1-(2-chlorophenyl)-N-methyl-(1-methylpropyl)-3-isoquinoline carboxamide, 1a (Chart 1),¹ is the major prototypical high-affinity ligand for the 18 kDa translocator protein (TSPO), formerly known as the peripheral benzodiazepine receptor. TSPO was first discovered as a result of its ability to bind diazepam² and was later distinguished from central benzodiazepine receptors by location, function, structure, and pharmacology.³⁻⁵ Several functions have been postulated for TSPO but perhaps the most evidence-based is as a mitochondrial membrane-based transporter, channel, or exchanger for cholesterol.⁶ Besides isoquinoline carboxamides⁷ and certain benzodiazepines (e.g., 2; Ro-5-4864, 4′-chlorodiazepam; Chart 1), TSPO also binds with high affinity to ligands belonging to many other structural classes^8,9 including quinoline carboxamides,¹⁰ pyrazolopyrimide acetamides,¹¹ 2-arylindole-3-acetamides,¹²N,N-dialkyl-2-phenylindol-3-ylglyoxylamides,¹³ and aryloxyanilides¹⁴⁻¹⁶ (Chart 1). High-affinity ligands from these and other structural classes invariably feature a single tertiary amido group.^8,9

Chart 1. Some High Affinity TSPO Ligands from Different Structural Classesa.

^a PK 11195 (1a; an isoquinoline carboxamide); Ro 5-4864 (2; a benzodiazepine); PBR 28 (3; an aryloxyanilide); FGIN 1 (4; an indoleacetamide); DPA-713 (5; a pyrazalopyrimidine); IGA-1 (6; an N¹-methyl-2-phenylindol-3-ylglyoxylamide).

TSPO is now implicated in various neuropsychiatric conditions, especially anxiety, and is expressed heavily in microglia in response to various inflammatory conditions in brain and periphery.^17,18 TSPO has therefore become a significant target for CNS drug development.⁹ Furthermore, radioligands¹⁹⁻²¹ for the imaging of TSPO in vivo are keenly pursued as potential biomarkers of inflammatory conditions.^22,23 These considerations drive the efforts to discover new and selective high-affinity TSPO ligands as potential new CNS drugs or improved imaging radioligands.

The interaction of 1a and other ligands with TSPO has been modeled extensively.²⁴⁻²⁷ From these studies, it is clear that the amido carbonyl group of 1a plays a critical role in its binding to TSPO, perhaps by forming a directional hydrogen bond within a generally lipophilic binding site. Hence, precise knowledge of the position and orientation of the carbonyl group of 1a in solution and when bound to TSPO can be informative about the topography of the TSPO binding site, and also aid in future ligand design. Earlier studies have not considered the possible existence of stable rotamers of 1a in solution (Figure 1), nor considered the possible roles of such rotamers in the binding of 1a to TSPO. Here, we performed dynamic NMR spectroscopy and quantum chemical studies on 1a that reveal four stable rotamers for 1a in solution that are due to amide bond and chlorophenyl group rotation. Our findings lead us to suggest that TSPO prefers to bind amide ligands as E rotamers and will inform future TSPO ligand and radiotracer design.

Figure 1.

Z/E isomerization of 1a through rotation of the dihedral angle of CH₃–N–C=O (ϕ₁); ϕ₂, ϕ₃, and ϕ₄ are the respective dihedral angles for CH₃–CH–N–CO, O=C–C3–C4, and C2′–C1′–C1–N.

Results and Discussion

Characterization of the Isomers of 1a

¹H NMR spectroscopy (400 MHz) of 1a in CDCl₃ (Figure S1, Supporting Information) or d₆-DMSO (Figure S2) at room temperature revealed the presence of amide bond rotamers. Signals for the N-methyl group protons in CDCl₃ are well separated for each rotamer (δ 2.90 and 2.98), as are those for the s-butyl group protons (Table 1). The signals for the s-butyl CH₂and CHCH₃ protons appear at lower field for the major rotamer than for the minor rotamer. However, signals for the s-butyl CH and CH₂CH₃ protons in the rotamers are reversed, with that at higher field belonging to the major rotamer. This reversal may be attributed to the magnetic anisotropy of the carbonyl bond.²⁸ For each pair of signals arising from chemically equivalent protons, the ratio of the integral for the major rotamer to that of the minor rotamer is in the range 1.8–2.0, indicating that the major rotamer is more stable than the minor rotamer in organic solution by about 0.4 kcal/mol. In addition, the ¹H NMR (400 MHz) spectrum of 1a in CDCl₃ at room temperature (Figure S1, Supporting Information; Table 1) clearly revealed an interesting unexpected feature for the signals of the N-methyl protons in the E and Z rotamers that was not seen in the spectra acquired in d₆-DMSO (Figure S2, Supporting Information). Namely, each signal appeared not as the expected singlet but as an almost completely resolved pair of peaks with small separation (<0.015 ppm) and unequal height, with that at lower field showing lower intensity. We suspected the existence of these paired peaks as being due to 2-chlorophenyl group rotamers.

Table 1. Theoretical and Experimental ¹H-NMR for 1a in Chloroform.

	chemical shift (δ)
	theory		experimental
signal	E1	Z1	E	Z
CH2CH3	0.34	0.75	0.79	0.99
	0.77	0.89	(2 overlapping t, J = 7.08 Hz)	(2 merging t, J = 7.38 Hz)
	0.85	1.05
CHCH3	0.78	1.02	1.22 (d, J =6.55 Hz)	1.19 (d, J = 6.34 Hz)
	1.35	1.22
	1.36	1.31	1.24 (d, J = 6.67 Hz)
CH2CH3	1.37	1.62	1.66 (m)	1.40 (m)
	1.74	1.72
NCH3	2.80	2.78	2.98 (d)	2.90 (d)
	2.82	2.85
	3.38	3.08
CH	4.40 (4.40)a	5.32 (5.39)a	3.89 (m)	4.80 (m)
ArH6′	7.93 (7.87)a	7.85 (7.95)a	7.66 (m)
ArH3′	8.13 (8.13)a	8.09 (8.16)a	7.37 (m)
ArH8	8.12 (8.0)a	8.10 (8.06)a	7.40 (m)
ArH5′	7.87 (7.83)a	7.90 (7.84)a	7.57 (m)
ArH4′	7.96 (7.85)a	7.96 (7.99)a	7.40 (m)
ArH7	8.15 (8.03)a	8.06 (8.08)a	7.57 (m)
ArH6	8.25 (8.23)a	8.21 (8.21)a	7.72 (m)
ArH5	8.56 (8.60)a	8.54 (8.55)a	7.96 (d, J = 8.33 Hz)
ArH4	8.78 (8.87)a	8.74 (8.67)a	8.01 (d)	8.05 (d)

^aValues in parentheses are for the respective E₂ and Z₂ isomers.

The ¹³C NMR spectrum of 1a in CDCl₃ at 24 °C (Figure S3, Supporting Information) also showed pairs of prominent signals for each alkyl carbon that are well separated at upper field with an intensity ratio comparable to that seen for the ¹H signals. For example, the intensity of the doublet signal at δ 55.75 and 55.57 is about 2-fold higher than the one at δ 50.58 and 50.38, suggesting that this pair of doubled signals, with a separation of more than 5 ppm, arises from amide bond rotamers in solution. Furthermore, the fact that each signal appeared as a very narrow doublet (separation ≤0.2 ppm), as seen in the signals of the N-methyl protons, again indicated that another pair of rotamers exists for each of the amide bond isomers.

While ¹H/¹³C COSY NMR experiments (Figure S4, Supporting Information) allowed correlation of the ¹H and ¹³C signals (e.g., those for C and H in the CH group of the s-butyl group) in the same rotamer, they cannot identify the rotamers or their geometries. Accordingly, we resorted to quantum mechanical calculations, utilizing density functional theory (DFT) at the level of B3LYP/6-311+G(2d,p) in the reaction field of chloroform, to obtain the structures and energetics of 1a rotamers as well as estimates of their ¹H and ¹³C NMR chemical shifts.

Figure 2 shows four fully geometry-optimized isomers of 1a, which we dub Z₁, Z₂, E₁, and E₂. The Z₁ isomer was modeled after the reported tetrameric X-ray structure of 1a showing a highly disordered s-butyl region,²⁹ whereas E₁ was obtained by rotating the N-methyl group of the Z₁ rotamer with respect to the amide C–N bond. Both Z₁ and E₁ display an exo orientation of the carbonyl group to the isoquinolinyl ring and a near orthogonality between this ring and the chlorophenyl group. The optimized Z₁ structure closely resembles the X-ray structure of higher occupancy fragment with a heavy atom root-mean-square deviation (rmsd) of 0.36 Å. When omitting consideration of the 2-chlorophenyl group, the rmsd decreased to 0.16 Å, showing that the rmsd of 0.36 Å is largely due to the difference in ϕ₄ [−102.9° (calculated) vs −78.2° (X-ray)]. The other calculated dihedral angles defined in Figure 1, [i.e., ϕ₁ (−168.8°), ϕ₂ (−100.6°), and ϕ₃ (48.5°)] are more similar to those of the X-ray structure [ϕ₁ (−162.8°), ϕ₂ (−95.4°), and ϕ₃ (56.4°)]. The X-ray structure also depicts a conformer of Z₁ with dihedral angle, ϕ₃ = −56.4°. This conformer, having a counter-clockwise orientation of the amide plane relative to the isoquinoline ring of Z₁, is discussed later in more detail.

Figure 2.

Geometry-optimized Z and E isomers of 1a at the level of B3LYP/6-311+G(2d,p) in the solvent reaction field of chloroform. Values in parentheses represent the total electronic energy and the Gibbs free energy at 298.15 K, respectively, relative to those of the E₁ isomer. The amide bond isomerization, Z₁ to E₁ or Z₂ to E₂, was done by varying ϕ₁ centered on the C–N bond; conversion of Z₁ into Z₂ or E₁ into E₂ was done by varying ϕ₄ centered on the C1–C1′ bond. Dashed lines indicate the distances between the CH and C6′H protons. Atoms are colored as follows: white, hydrogen; green, carbon; blue, nitrogen; red, oxygen; violet, chlorine.

When the chlorophenyl groups of Z₁ and E₁ were rotated with respect to their C1–C1′ bonds by about 180°, calculations revealed another pair of stable rotamers, Z₂ and E₂. In terms of the total electronic energy in the reaction field of chloroform, E₁ appears most stable among the four isomers. When the zero point correction and thermal free energy at 298.15 K were included, Z₁ and E₁ became isoenergetic and more stable than Z₂ and E₂ by 0.4 and 0.6 kcal/mol, respectively. The extra stability of both Z₁ and E₁ likely arises from lower steric repulsion between the amide oxygen pointing out of the plane and the C2′–Cl pointing into the plane (Figure 2). However, these calculated energy differences are too close to assign the observed more intense NMR peak of a pair to either the Z₁ or E₁ isomer. We therefore calculated NMR chemical shifts for each rotamer since such calculations are known to be very useful for assigning experimental ¹H and ¹³C peaks to specific isomers.³⁰

Figure 3 represents the four rotamers of 1a with their calculated ¹³C chemical shifts. For the rotamer identification, we selected four carbons (i.e., N-CH₃, CHCH₃, CHCH₃, and C=O) in the amide region because the differential magnetic shieldings of these carbons are stronger than for others. For example, from the calculations, the N-CH₃ of E₁ (δ 29.20) syn to oxygen is much more strongly shielded than that of Z₁ (δ 32.97) anti to oxygen. Also, both CHCH₃ (δ 57.11) and CHCH₃ (δ 19.01) of Z₁syn to oxygen are better shielded than the respective anti carbons of E₁ (δ 65.28, δ 20.34). The shielding and deshielding of the alkyl carbons of the E and Z isomers are consistent with previous ¹³C NMR studies,³¹⁻³⁴ which have shown that alkyl carbon atoms syn to amide oxygen are better shielded than the corresponding anti carbons. In particular, the differential shielding of the N-methyl carbon was attributed either to the electric field caused by the carbonyl group^33,34 or to the paramagnetic contribution of the N–C bond of the N-methyl group.³¹ Besides the syn/anti effect in upper field, our calculations further indicate that the carbonyl carbon in Z₁ (δ 179.54) is better shielded than in E₁ (δ 180.68), which can serve as a low field marker.

Figure 3.

Calculated ¹³C chemical shifts for four isomers of 1a at the level of B3LYP/6-311+G(2d,p) in the solvent reaction field of chloroform.

Whereas amide bond rotation results in sizable chemical shift differences between the alkyl carbons of Z₁ and E₁, much smaller differences are seen for chlorophenyl group rotation between Z₁ and Z₂ or between E₁ and E₂ (Figure 3), in good agreement with the narrow doublet of each alkyl carbon signal observed experimentally (Figure S3, Supporting Information). For instance, the s-butyl CH of Z₂ (δ 57.14) is very slightly deshielded upon rotation of the chlorophenyl group of Z₁(δ 57.11). Accordingly, the observed doublet at δ 50.38 and 50.58 can be assigned to the CH of Z₁ and Z₂, respectively. Similarly, the doublets seen at δ 55.75 and 55.57 were assigned, respectively, to the calculated CH of E₁ (δ 65.28) and E₂(δ 64.30). The upfield experimental ¹³C peaks were assigned by using the calculated chemical shift caused by amide bond rotation as a major marker and those of the chlorophenyl group rotation for fine-tuning (Table S1, Supporting Information). Figure 4 depicts the experimental and calculated chemical shifts for the alkyl carbons of 1a without the resolution of their chlorophenyl group rotamers. This assignment clearly demonstrates that (i) the syn/anti effect is stronger for CH and CH₃ directly bonded to the amide nitrogen and then gets progressively weaker for the more distal carbon atoms and (ii) that for any given pair, the experimental peak intensity assigned to E isomer is stronger.

Figure 4.

Pictorial comparison of the calculated and experimental ¹³C NMR spectra of Z and E of isomers of 1a at upper field. Tall and short lines in the experimental spectrum indicate the respective major and minor signals, for each carbon.

Whereas use of calculated and experimental ¹³C signals to identify E and Z isomers can be rather straightforward, use of ¹H signals is less certain due to the narrower span of chemical shifts. Moreover, for each methyl group the calculations provide three distinct signals, one for each methyl proton, whereas experimental ¹H NMR gives only a single signal at room temperature. This renders any comparison between theory and experiment more challenging. For 1a, the s-butyl CH proton turned out to be useful. The calculated chemical shifts indicate that the CH proton in E₁ (δ 4.40) and E₂ (δ 4.40) is better shielded than in Z₁ (δ 5.32) and Z₂ (δ 5.39), and this agrees well with the experiment (E rotamer, δ 3.89; Z rotamer, δ 4.80) based on relative peak intensity (Table 1). In general, protons syn to the amide oxygen are better shielded than those anti.³⁵ However, when a methine (CH) proton in the plane of the amide group is conformationally restricted, such shielding is reversed.³⁶ The calculated energy barrier for the rotation of ϕ₂ of Z₁ is about 11 kcal/mol, thereby confirming the constrained rotation of the s-butyl group with respect to the N–CH bond. The better experimental shielding of CHCH₃ protons in Z(δ 1.19) than in E (δ 1.22) and also of CH₂CH₃ protons in E (δ 0.79) than in Z (δ 0.99) can be attributed to the syn effect to the amide oxygen as seen in Figure 2. Our experimental observation that the N-CH₃ is more strongly shielded in Z(δ 2.90) than in E (δ 2.98) is noteworthy in view of numerous reports that freely rotating N-methyl protons syn to an amide oxygen are better shielded.^36,37 In accord with experiment, calculation also shows the calculated N-CH₃ peaks are on average better shielded in Z₁ (δ 2.90) than in E₁ (δ 3.00). The geometry-optimized structures of 1a (Figure 2) show that the N-CH₃ of Z₁ but not of E₁ is in the diamagnetic region of the isoquinolinyl ring current field and that this ring current likely causes the reversal of the N-CH₃ peaks.^36,38

Both ¹H and ¹³C peak intensity invariably favored assignment of E as the major rotamer of 1a in solution. This does not fully agree with our calculations, which indicate that Z₁ and E₁ are isoenergetic. The only unambiguous experimental method for resonance assignment in tertiary amides is through observation of a clear nuclear Overhauser effect (NOE).³⁹ Accordingly, we searched for interactions between the s-butyl CH proton signal and those of the chlorophenyl group protons with NOE spectroscopy of 1a in CDCl₃. Only the major rotamer showed a positive effect (Figure S5, Supporting Information), and therefore, we unambiguously assign this rotamer as E and the minor rotamer as Z. This result is consistent with the optimized structures of 1a in Figure 2, which show that the s-butyl CH protons of E₂ and Z₁ are positioned at 3.60 Å and 6.66 Å away from their corresponding chlorophenyl group C6-H protons, respectively, whereas the s-butyl CH protons of E₁ and Z₂ are positioned at 5.32 Å and 6.91 Å away from their corresponding protons.

The preference for 1a to exist as an E rotamer in organic solution contrasts with the existence of 1a as a Z rotamer in the crystalline state.²⁹ Our calculations indicate that the Z₁ rotamer with the chlorine pointing into the plane (Figure 2) is more stable by 0.4 kcal/mol at room temperature than Z₂ with the chlorine pointing out of the plane. This finding agrees well with the X-ray structure that has shown 94% site-occupation for the chlorine at 2′-position corresponding to Z₁ in Figure 2. Nevertheless, it is puzzling why the E isomers become more stable by 0.4 kcal/mol than the Z isomers in organic solution. One plausible explanation is that the solvent reaction field, such as the polarizable continuum model (PCM) implemented in quantum chemistry software, may not fully account for the solvation of Z and E isomers by CHCl₃ or DMSO. Accurate estimation of the interaction energy between solute and solvent requires the explicit structure of the solvent, at least at the level of the first solvation shell.⁴⁰ Thus, while good at approximating the electrostatic interaction between solute and bulk solvent, the reaction field itself may not predict the solvent induced stability of the E over the Z isomer. In this case, the specific short-range solute–solvent interactions may be playing a critical role.

Characterization of the Isomers of 1b

In contrast to the findings on 1a, both the ¹H (Figure S6, Supporting Information)- and ¹³C NMR (Figure S7, Supporting Information) spectra of the corresponding N-desmethyl secondary amide 1b at 24 °C in CDCl₃ showed no duplications of signals attributable to amide bond rotamers. The two possibilities are that the rotamers are rapidly interconverting on the NMR time-scale or that a single rotamer predominates.³⁵ Generally, most secondary amides are found to exist as Z rotamers only.³⁵ The calculated energetics of 1b indicate that the geometry-optimized Z rotamer (Figure 5) is 5.8 kcal/mol more stable than the E rotamer, which implies that more than 99.99% of 1b exists in Z form in organic solution at room temperature.

Figure 5.

Geometry-optimized Z and E isomers of 1b. Dashed lines indicate the distance between the hydrogens of CH and C6′H and also between the NH and the nitrogen of the isoquinoline ring. The value in parentheses represents the Gibbs free energy at 298.15 K with respect to that of the Z isomer.

The large stability difference between the rotamers of 1b arises from the spatial orientation of the amido hydrogen. In the Z rotamer, the isoquinolinyl nitrogen forms a hydrogen bond with the amide hydrogen. As a result, both the isoquinolinyl ring and the amide group are on the same plane (ϕ₃ = 0.3°). By contrast, in the E rotamer, there is steric repulsion between the isoquinolinyl ring and the s-butyl group, resulting in a sizable distortion between the two planes (ϕ₃ = 43.1°).

An NOE experiment revealed no interaction between the signal for the s-butyl CH proton and those of the protons in the chlorophenyl ring. This is consistent with the optimized geometry of Z (Figure 5) showing its s-butyl CH proton 6.70 Å away from the chlorophenyl group C6′H. The s-butyl CH proton of E can interact with the chlorophenyl group C6′H upon chlorophenyl group rotation. However, the NOE signal was not detected due to the thermochemical instability of the E rotamer. Hence, the Z configuration was assigned to 1b. The ¹H NMR spectrum calculated for the Z rotamer of 1b alone quite accurately predicts the key features of the experimental ¹H NMR spectrum (Table 2). Notably, the calculation suggests that the chemical shifts for the NH proton differ significantly between rotamers (Z, δ 8.74; E, δ 6.69), and this feature most distinguishes the spectra; the amide hydrogen in the Z isomer of 1banti to the oxygen is much more deshielded than in the E isomer. The signal for the NH proton in the experimental spectrum falls wholly between δ 8.04 and 8.06 and therefore, within the magnitude of computational error, is consistent with 1b existing exclusively as the Z rotamer. As in the case of 1a, the s-butyl CH proton of E (δ = 4.36) was calculated to be better shielded than that of Z (δ 4.98); experimentally, this proton was observed at δ 4.16.

Table 2. Theoretical and Experimental ¹H NMR Data for 1b in Chloroform.

	chemical shift (δ)
signal	E (theory)	Z (theory)	experiment
CH2CH3	0.33	0.79	0.964 (t, J = 6.99 Hz)
	0.87	0.93
	1.05	1.05
CHCH3	0.95	0.95	1.27 (d, J = 6.62 Hz)
	0.97	1.18
	1.59	1.40
CH2CH3	1.61	1.80	1.61 (m)
	1.33	1.46
CH	4.36	4.98	4.16 (m)
NH	6.69	8.74	8.04 (m)
ArH6′	7.89	7.92	7.68 (m)
ArH3′	8.18	8.19	7.50 (m, 3H)
ArH8	8.06	8.07
ArH4′	7.95	8.06
ArH5′	7.85	7.89	7.58 (m, 2H)
ArH7	8.13	8.09
ArH6	8.25	8.22	7.75 (m)
ArH5	8.60	8.63	8.05 (d, J = 8.20 Hz)
ArH4	8.92	9.36	8.67 (d

In line with the ¹H NMR study, the observed ¹³C NMR spectrum of 1b (Figure S7, Supporting Information) does not show a pair of amide isomers. As shown in Figure 6, the calculated ¹³C chemical shifts of the alkyl carbons, such as CHCH₃, CHCH₃, and CH₂CH₃, exhibit a pattern similar to that seen for 1a in that the carbons syn to oxygen are better shielded than the corresponding anti carbons. For example, the s-butyl CH of Z (δ 52.52) syn to oxygen is 6.68 ppm better shielded than that of E (δ 59.20). Accordingly, the experimental chemical shift (δ 44.92) can be safely assigned to the CH of the Z isomer (Table S2, Supporting Information). Most of these alkyl carbon peaks also exist as a doublet separated by less than 0.07 ppm that can be attributed to chlorophenyl group rotation.

Figure 6.

Calculated ¹³C chemical shifts for the Z and E isomers of 1b.

Besides the alkyl-carbons, our calculations suggest that the carbonyl carbon of Z in 1b (δ 171.69) is unusually more strongly shielded than that of E (δ 179.55). Moreover, the chemical shift of the C=O of desmethyl-E is comparable to that of Z₁ (δ 179.54) and E₁ (δ 180.68) of 1a. In line with the calculations, the experimental chemical shift for C=O in 1b(δ 162.22) was also significantly shifted upfield with respect to that in 1a (δ 168.12 and δ 168.38). A notable geometrical feature that might provide a plausible rationale for the observed chemical shifts for C=O is the distortion between the planes of the isoquinolinyl ring and the amide group, as represented by ϕ₃. While the Z form of 1b has essentially no distortion (ϕ₃ = 0.3°), all others have a sizable distortion from planarity, i.e., E of 1b (ϕ₃ = 43.1°), Z₁ of 1a (ϕ₃ = 48.5°), and E₁ of 1a (ϕ₃ = 49.3°). This has led us to hypothesize that the planarity between the isoquinolinyl ring and the amide group gives rise to the strong shielding of the carbonyl carbon of 1b. A plot of the calculated chemical shift for C=O as a function of the dihedral angle, ϕ₃, in 1b (Figure 7) indeed demonstrated that shielding increases as ϕ₃ changes from ±40° to 0°. The exact nature of the strong magnetic anisotropy of the isoquinolinyl ring on the C=O of desmethyl-Z is being investigated. Nonetheless, this unusual shielding can be a useful prominent marker for the Z rotamer of 1b in particular as well as for other benzamides.

Figure 7.

Calculated ¹³C-chemical shifts of the carbonyl carbon as a function of ϕ₃ in 1b.

Dynamics of E and Z Rotamer Interconversion in 1a

Having successfully characterized the E and Z isomers of 1a and 1b in solution via experimental and calculated ¹H/¹³C chemical shifts, we also investigated isomer interconversion by utilizing dynamic NMR and quantum chemistry. At elevated temperatures, each duplicate ¹H NMR signal from the amide bond rotamers of 1a merges to give a single peak at a specific coalescence temperature, T_c. T_c values were measured on 1a in d₆-DMSO for the well-defined signal pairs from the N–CH₃, s-butyl CH, s-butyl CH₂CH₃, and C4H protons. Free energies for the rotation barriers Z to E as well as E to Z were then calculated by the method of Shanan-Atidi and Bar-Eli.⁴¹The values derived for these two barriers from each of the four pairs of ¹H NMR signals are in excellent agreement (17.1 to 17.8 kcal/mol; Table 3) and are in the range that is often observed for tertiary amides.³⁵ The E rotamer was found to be more stable than the Z rotamer by about 0.38–0.48 kcal/mol at T_c values of 57–104 °C.

Table 3. Experimentally Determined Energy Barriers for Amide Bond Rotation in 1a in d₆-DMSO^a.

	δ (ppm)					(kcal/mol)
1H NMR signal	Z	E	|Δν| (Hz)	ΔP (E – Z)	Tc (°C)	ΔG⧧E	ΔG⧧Z	ΔG⧧(E–Z)	kr (Hz)
N-CH3	2.77	2.85	29.8	0.288	71	17.8	17.4	0.38	66
s-Bu CH2CH3	0.90	0.69	80.7	0.286	85	17.8	17.5	0.39	179
s-Bu CH	4.61	3.72	357	0.294	104	17.7	17.3	0.41	793
C4-H	8.13	8.09	16.0	0.340	57	17.6	17.1	0.48	36

^aPreviously undefined column headings are defined as follows: Δν = difference in chemical shifts for E and Z¹H NMR signals. ΔP = difference in fractional populations between Z (P_Z) E (P_E) and rotamers. ΔG^⧧_Z = activation energy for rotation from Z rotamer. ΔG^⧧_E = activation energy for rotation from E rotamer. ΔG^⧧_(E–Z)= ΔG^⧧_E –ΔG^⧧_Z. k_r = rate of rotation of bond at stated coalescence temperature (T_c).

The rate (k_r) of E/Z rotation may be estimated at each coalescence temperature as (π/2^1/2)Δν, if the small differences in E and Z rotamer populations are neglected.⁴² These rates are listed in Table 3 at each of the four recorded coalescence temperatures. They are comparable to rates determined for several other tertiary aryl amides.⁴³ As the rate data for amide bond rotation in 1a were obtained at coalescence temperatures spread over 34 °C, it was possible to draw an Arrhenius plot of ln k_r versus 1/T (Figure S8, Supporting Information). Although, based on only 4 points, this plot was found to be highly linear (r² = 0.976) and allowed the Arrhenius activation energy (E_a) to be derived from its slope (−E_a/R). The value obtained for E_a is 16.5 kcal/mol, which is just 1 kcal/mol lower than the mean estimate of the free energy barrier to rotation (ΔG^⧧ = 17.5 kcal/mol) (Table 3). Since E_a = ΔH^⧧ + RT, the enthalpy of activation ΔH^⧧ may be estimated and then also the entropy of activation ΔS^⧧, according to ΔG^⧧ = ΔH^⧧ – TΔS^⧧. The values obtained were ΔH^⧧ = 15.9 kcal/mol, and ΔS^⧧ = −5.33 cal·°C^–1·mol^–1. The small negative value for ΔS^⧧ is in accord with constrained vibrational motions at the transition state and is also consistent with similar values reported for other amide bond rotations.⁴²⁻⁴⁴ Finally, the intercept of the Arrhenius plot at 1/T = 0 gives 12.40 as the value for the log₁₀ of the frequency factor (A), which is also in the range reported for amide bond rotations, such as those in N,N-dimethylformamide.^45,46 The Arrhenius plot permits estimation of the rate of rotation at any temperature of interest (e.g., k_r is estimated to be ∼1 Hz at 20 °C).

Dynamic NMR in CDCl₃ also showed coalescence of each of the two pairs of N-CH₃ signals from which the barrier for the chlorophenyl group rotation was readily estimated (Table 4). The rate of rotation at the measured coalescence temperatures (45 and 44 °C for Z and E rotamers, respectively) was estimated according to the expression k_r = (π/2^1/2)Δν to be of the order of 7 to 12 Hz. This estimate may be quite accurate because members of each rotamer pair have almost equal population. Scrutiny of the same ¹H-NMR spectra revealed that the two singlets for the C-4 aryl hydrogen were also split into two peaks of almost equal intensity. Each pair of peaks merged at coalescence temperatures of 55 and 57 °C for the Z and E rotamers, respectively (Table 4). The energy barriers for chlorophenyl group rotation (e.g., ΔG^⧧_A = 18.0 kcal/mol) obtained from the coalescence of the C4-H peaks agree exceptionally well with those determined from the respective N-methyl signals (e.g., ΔG^⧧_A = 18.1 kcal/mol). The ΔG^⧧_A of 18.0 kcal/mol for the E is 1.0 and 0.4 kcal/mol higher than that for the respective chlorophenyl group rotation of the Z and the amide bond rotation energy barrier (ΔG^⧧_E = 17.6 kcal/mol), suggesting that the interconversion of the four isomers shown in Figure 2 occurs over a similar time scale. Rates for chlorophenyl group rotation in the E and Z isomers were also estimated (Table 4). No attempt was made to obtain Arrhenius equation parameters from these data which are for only two close T_c values.

Table 4. Experimentally Determined Energy Barriers for 2-Chlorophenyl Group Rotation in 1a in CDCl₃.

	δ (ppm)					(kcal/mol)
1H NMR signal	signal A	signal B	|Δν| (Hz)	ΔP(A – B)	Tc (°C)	ΔG⧧A	ΔG⧧B	ΔG⧧(A – B)	kr (Hz)
N-CH3 (Z)	2.90	2.89	5.28	0.100	45	17.3	17.2	0.10	12
C4-H (Z)	8.04	8.07	15.5	0.077	55	17.1	17.0	0.08	35
N-CH3 (E)	2.98	2.98	3.00	0.270	44	18.1	17.5	0.63	7
C4-H (E)	8.01	7.99	5.40	0.143	57	18.0	17.8	0.21	12

Potential Energy Surfaces (PESs) for the Interconversion of Z and E Rotamers of 1a and 1b

In view of our ¹H NMR observations, revealing four quite stable rotamers of 1a, but not of 1b, the PES for the conversion of both 1a and 1b was constructed in order to gain further insight into kinetic processes based on the structure and energetics of their ground and transition states.

We obtained the PES of 1a in the gaseous phase at the level of B3LYP/6-31G* by varying the ϕ₁ (C_Me–N-C=O) from −164.7° to 175.3° with an increment of 10° while relaxing the rest of the structure (Figure 8, panel A). This PES shows a steady rise in energy as the amide central C–N bond of Z₁ weakens. Upon passing over the first energy barrier, another stable conformer (E₁′) forms, which resembles E, but has the opposite orientation of the amide group with respect to the isoquinoline ring. Further change in ϕ₁ converts E₁′ back to Z₁ after passing the second energy barrier. Transition state (TS) geometry optimization, at the level of B3LYP/6-31G* in the solvent reaction field of chloroform, at these two high energy points gave the two TSs with energy barriers of 18.7 and 20.1 kcal/mol at ϕ₁ = −64.8° and 105.3°, respectively. Both TSs are characterized by a weakened C–N bond (1.43 Å) relative to that in the ground states (1.36 Å) and also by a single imaginary vibrational frequency (Figure S9, Supporting Information). The fully geometry-optimized conformer E₁′ with the ϕ₃ = −36.0° is 1.1 kcal/mol less stable than E₁ with the ϕ₃ = 40.2°. The calculated activation free energy (ΔG^⧧) going from E₁ to E₁′ is 4.1 kcal/mol, rendering these forms unresolvable by NMR at room temperature. The calculated ΔG^⧧ of 18.7 kcal/mol from Z₁ to E₁′ in 1a (Table 5) is comparable to 17.4 kcal/mol as determined with ¹H NMR for the coalescence of the N-CH₃ signal in d₆-DMSO. Calculation also gave estimates of ΔH^⧧ (16.5 kcal/mol) and ΔS^⧧ (−7.21 cal·K^–1·mol^–1) in good agreement with our experimental estimates (ΔH^⧧ = 15.9 kcal/mol and ΔS^⧧ = −5.33 cal·°C^–1·mol^–1).

Figure 8.

Potential energy versus amide group dihedral angle (ϕ₁) in 1a (panel A) and in 1b (panel B). Ground and transition states before full geometry optimization are indicated by their labels.

Table 5. Energies and Geometrical Parameters of 1a and 1b Optimized at the B3LYP/6-31G* Level in the Solvent Reaction Field of Chloroform with the PCM Model.

structure	state	ϕ1 (°)	N–C(O) bond length (Å)	electronic energy + thermal enthalpya(kcal/mol)	electronic energy + thermal free energya(kcal/mol)	entropy(cal·mol–1·K–1)
1a	Z1	–165.2	1.36	0	0	165
1a	TS1	–64.8	1.43	16.5	18.7	158
1a	TS2	105.3	1.43	18.6	20.1	160
1a	E1′	–0.7	1.36	1.2	1.3	165
1b	Z	–176.7	1.35	0	0	159
1b	TS1	–62.8	1.43	23.8	25.7	153
1b	TS2	108.2	1.43	24.0	25.4	154
1b	E	2.1	1.36	7.1	7.4	158

^aRelative energetics at 298.15 K.

The PES for the conversion of Z to E for 1b is very similar to that of 1a in that it has two minimal energy conformations that are distinguished as E and Z rotamers with H–N–C=O (ϕ₁) torsion angles of 2.1° and −176.7°, respectively (Figure 8, panel B). However, the barrier for Z to E rotation in this secondary amide is 6.7 kcal/mol greater than that in 1a due to the extra stabilization of the Z form, as discussed earlier. The greater stability of the Z rotamer is also reflected a slightly shorter amide bond length ((N–CO), 1.35 Å) than in the E rotamer (1.36 Å) (Table 5).

PES for the Rotation of the Chlorophenyl Ring

Figure 9 depicts the PES for varying chlorophenyl group rotation (ϕ₄) in the Z₁ isomer of 1a. The respective first and second energy barriers are 16.5 and 19.7 kcal/mol in the gaseous phase without zero-point and thermal contribution. Upon subjecting 1a to further TS geometry optimization in the solvent reaction field of chloroform, the ΔG^⧧ for the first and second TSs were calculated to be 19.8 and 24.1 kcal/mol, respectively. The larger second TS energy barrier mainly arises from the stronger steric repulsion between the Cl atom and the C8-H atom of the isoquinolinyl moiety (Figure S10, Supporting Information). When the calculation was performed on deschloro-1a (7), the barrier for the first TS was reduced to 7.8 kcal/mol. This indicates that the steric repulsion between the nonbonding sp² electrons of the isoquinoline nitrogen and the chlorine atom at the first TS in 1a amounts to about 12 kcal/mol and thus hinders the chlorophenyl group rotation significantly. A similar trend was observed for the chlorophenyl group rotation in the E₁ isomer, resulting in the first TS energy barrier of 18.9 kcal/mol (Table 6). Interestingly, the N-desmethyl compound 1b has a comparable energy barrier (20.0 kcal/mol), indicating that the N-methyl group does not much influence the rotational barrier of the chlorophenyl ring. The calculated values are 2.5 and 0.8 kcal/mol higher, respectively, than the experimental ΔG^⧧_A values for the N-CH₃ (Z) and N-CH₃ (E). The discrepancy of 2.5 kcal/mol for the chlorophenyl group rotation of Z₁ might be attributed to the lack of explicit solvation in calculations or other reaction paths with a lower energy barrier that were not explored in the present study. The latter appears to be plausible because the calculated ΔG^⧧ was reduced to 18.6 kcal/mol when another conformer of Z₁′ with the ϕ₃ = −37.1° was employed (Table 6).

Figure 9.

Potential energy versus chlorophenyl group rotation in the Z₁ isomer of 1a. Ground and transition states before full geometry optimization are indicated by their labels.

Table 6. Chlorophenyl Group Rotational Energy Barriers for 1a, Deschloro-1a (7), and 1b Obtained at the B3LYP/6-31G* Level in the Solvent Reaction Field of Chloroform with the PCM Model.

compd	state	ϕ4 (°)	electronic energy + thermal enthalpya(kcal/mol)	electronic energy + thermal free energya(kcal/mol)	entropy(cal.mol–1·K–1)
1a	Z1	–107.6	0	0	165
1a	TS	11.5	17.0	19.8	156
1a	Z2	82.0	0.4	–0.1	167
1a	E1	–105.9	0	0	164
1a	TS	11.6	17.3	18.9	159
1a	E2	71.8	0.5	0.6	164
1a	Z1′	–106.7	0	0	163
1a	TS	–11.5	16.2	18.6	156
1a	Z2′	106.6	–0.4	–0.3	163
7	Z	–129.4	0	0	156
7	TS	5.9	6.9	7.8	153
1b	Z	–106.7	0	0	159
1b	TS	11.1	17.7	20.0	151

^aRelative energetics at 298.15 K.

Additional Conformers of 1a Arising from the Rotation of ϕ₃ and Interconversion Paths

Besides the four stable isomers that are interconvertible via the rotation of ϕ₁ or ϕ₄ at room temperature, four additional isomers Z₁′, Z₂′, E₁′, and E₂′ (Figure 10) were obtained through the rotation of ϕ₃. Interestingly, these two sets of four isomers are conformationally diastereoisomeric since the counter-clockwise rotation of ϕ₃ with respect to the C3–CO bond results in almost an inversion of the amide group. For example, Z₁′ with (ϕ₃ = −50.3° and ϕ₄ = −99.9°) is diastereomeric to Z₂ (ϕ₃ = 48.0° and ϕ₄ = 104.4°). The Z₁′ form was observed in the X-ray structure of 1a and is calculated to be less stable by 0.7 kcal/mol than Z₁ at the level of B3LYP/6-311+G(2d,p). However, the calculated ΔG^⧧ between Z₁ and Z₁′ is only 3.8 kcal/mol, making these rotamers unresolvable at room temperature. When such rapid interconversion occurs in solution, calculated chemical shifts with a weighted average may compare better with experimental NMR data. However, this will require an accurate energy calculation together with an explicit solvent model. Our calculated NMR chemical shifts for Z₁, Z₂, E₁, and E₂ are, nonetheless, very compatible with experimental findings, and thus, we made no attempts to calculate the weighted NMR chemical shifts. The calculated ¹³C chemical shifts for Z₁′, Z₂′, E₁′, and E₂′ (Figure S11, Supporting Information) exhibit a pattern similar to those in Figure 2 and are provided for comparison.

Figure 10.

Possible interconversion pathways for 1a.

Figure 10 also depicts a number of possible interconversion paths among the eight rotamers of 1a. For example, the conversion of Z₁ to E₂ may well proceed via the path of Z₁-E₁-E₂ or Z₁-Z₂-E₂. A valid question is then whether this conversion would occur in a concerted or discrete manner. Dynamic NMR clearly showed two distinct coalescent processes associated with the amide bond as well as chlorophenyl group rotation. In support of the dynamic NMR finding, the calculated TS structures (Figures S9 and S10, Supporting Information) do not show a sizable steric clash between the alkyl portion and the chlorophenyl group of 1a. All these data suggest that the conversion of Z₁ to E₂ or vice versa occurs largely in a discrete manner.

TSPO Preferentially Binds the E Form?

The fact that 1a exists as interconvertible rotamers that are energetically similar raises an interesting question as to which rotamer is better recognized by TSPO. One possibility is that TSPO prefers to bind amide rotamers with an E configuration. As is the case for 1b, the E configuration may be considered generally to be absent in secondary amides.³⁵ A preference for TSPO to bind E rotamers would be consistent with the observation that TSPO invariably binds tertiary amides much more strongly than their corresponding secondary amides, as illustrated by several examples (Table S3, Supporting Information).^25,47 Thus, for the tertiary/secondary amide pair, 1a/1b, the difference in energy of binding to TSPO estimated from the IC₅₀ values is about 4.4 kcal/mol. [The difference in TSPO binding energy (ΔB_E) for 1a and 1b may be estimated according to ΔB_E = RT[ln(^1bIC₅₀/^1aIC₅₀)], where R is the gas constant (1.986 cal °C/mol),T is the absolute temperature (4 °C = 277.15 K), and ^1aIC₅₀ and ^1bIC₅₀ are the IC₅₀ values for the binding of 1a and 1b, respectively, to TSPO.] The approximate energy parity of the Z and E rotamers of 1a, compared to the large energy disparity (5.8 kcal/mol) between the amide bond rotamers of 1b, might easily account for this binding energy difference. Moreover, in a few known cases, where a quinoline or isoquinoline tertiary amide ligand has been constrained to the Z form, the affinity for TSPO is low by several thousand-fold relative to the structurally closest unconstrained tertiary amide ligand (c.f., 8 with 1c, Table S3, Supporting Information).^25,27 Energetically favorable freedom to rotate to the E form may therefore be a prerequisite for high affinity binding of ligands based on isoquinoline/quinoline amides to TSPO.

Conclusions

The solution structures of 1a and 1b arising from the amide bond and chlorophenyl group rotation were fully characterized. While 1a exists as both Z and E rotamers in solution, 1b exists exclusively in Z form. Our experimental finding that the E isomer of 1a is more stable than the Z isomer in solution together with our in-depth study of their interconversion should better inform future studies aimed at acquiring a deeper understanding of TSPO–ligand interactions and new TSPO ligand discovery for drug development and molecular imaging.

Methods

Materials

Compound 1a (racemate) was purchased from Sigma-Aldrich (USA) and (R)-N-desmethyl-1a (1b) from ABX (Germany).

Spectroscopy

¹H NMR spectra (400.13 MHz) and ¹³C NMR (110 MHz) were recorded with an Avance spectrometer (Bruker) in the solvent indicated with tetramethysilane (TMS) as internal standard. The values of chemical shifts are expressed in ppm downfield from the TMS signal, and coupling constants J are expressed in Hz.

¹H NMR of 1a

The ¹H NMR spectrum was obtained on 1a(26.7 mg/mL) in CDCl₃ at 24 °C. COSY 90, NOESY, HMBC, HMQC, and NOE spectra were also obtained to assist with signal assignment.

Dynamic ¹H NMR of 1a

Coalescence temperatures (T_c) associated with amide bond rotation in 1a were determined by raising the temperature at which ¹H NMR (d₆-DMSO) spectra were obtained in 5 °C increments from 20 °C until proton signals for (i) N-methyl, (ii) s-butyl CH₂CH₃, (iii) s-butyl CH, and (iv) C4-H were each found to merge. Further spectra were then acquired to determine each of the four T_c values to within less than one °C.

Coalescence temperatures (T_c) for 2-chlorophenyl group rotation were determined by acquiring ¹H NMR (CDCl₃) spectra in 5 °C increments from −5 °C until the paired proton signals for N-methyl protons and C4–H proton in each of the E and Z rotamers were seen to coalesce. Further spectra were then acquired to determine T_c to within less than 1 °C.

Energy barriers to rotations in 1a were calculated from the acquired NMR data and T_c values according to the method of Shanan-Atidi and Bar-Ali.⁴¹ For more details, see Supporting Information.

¹³C NMR of 1a

The ¹³C NMR spectrum was obtained on 1a(26.7 mg/mL) in CDCl₃ at 24 °C.

¹H NMR of 1b

The ¹H NMR spectrum of 1b (20 mg/mL) was recorded in CDCl₃ at 24 °C. HETCOR (direct HC), HMBC, and H–H COSY spectra were also obtained to assist in assignment of signals. Spectra were also run in C₆D₆ to assist with signal assignment.

¹³C NMR of 1b

The ¹³C NMR spectrum was obtained on 1b(26.7 mg/mL) in CDCl₃ at 24 °C.

Quantum Chemistry

The geometries of the rotamers of 1a and 1b were optimized with density functional theory at the B3LYP/6-311+G(2d,p) level.⁴⁸ The PES for the rotation of the amide bond (ϕ₁) and of the chlorophenyl group (ϕ₄) were constructed by varying a torsional angle in increments of 10° while optimizing all other geometries in the gaseous phase at the level of B3LYP/6-31G*. The geometries of both E and Z rotamers as well as those of the TSs were further optimized utilizing the polarized continuum model with the UAKS parameter sets to incorporate the solvent (CHCl₃) effect. Each TS was confirmed by the existence of a single imaginary frequency.

Quantum Chemical Calculation of ¹H and ¹³C NMR Spectra

Shielding values (in ppm) in the solvent reaction field of chloroform at the level of B3LYP/6-311+G(2d,p) were calculated for all carbons and protons in each rotamer of 1a and 1b employing the gauge-independent atomic orbital method.⁴⁹¹H and ¹³C chemical shifts were obtained by subtracting the shielding value of each proton and carbon from 31.8885 and 184.0456, respectively, the calculated isotropic shielding tensor for the protons and carbons in TMS.

Glossary

Abbreviations

CIT

Center for Information Technology

CNS

central nervous system

DFT

density functional theory

COSY

correlated spectroscopy

DMSO

dimethyl sulfoxide

HETCOR

heteronuclear correlation spectroscopy

HMBC

heteronuclear multiple bond correlation

HMQC

heteronuclear multiple quantum coherence

NOE

nuclear Overhauser effect

NOESY

nuclear Overhauser effect spectroscopy

NIH

National Institutes of Health

NIMH

National Institute of Mental Health

NMR

nuclear magnetic resonance

PCM

polarizable continuum model

PES

potential energy surface

PET

positron emission tomography

PK 11195

1-(2-chlorophenyl)-N-methyl-N-(1-methylpropyl)-3-isoquinolinecarboxamide

rmsd

root-mean-square deviation

TMS

tetramethylsilane

TS

transition state

TSPO

translocator protein

Supporting Information Available

Binding assay methods and results; details of NMR and dynamic NMR of 1a and 1b; Arrhenius plot; and TS structures. This material is available free of charge via the Internet at http://pubs.acs.org.

Author Contributions

Y.-S.L. oversaw and performed quantum chemistry, interpreted data, and contributed to writing of the manuscript, especially parts concerned with quantum chemistry. F.G.S. and E.B. performed NMR spectroscopy and contributed to NMR data interpretation. V.W.P. initiated and directed the project, designed experiments, analyzed and interpreted data, wrote major parts of the paper, and composed the final draft.

This research was supported by the Intramural Research Program of the National Institutes of Health (CIT and NIMH).

The authors declare no competing financial interest.

Funding Statement

National Institutes of Health, United States