Subscriber access provided by University of Newcastle, Australia
Article
Development of a 13C NMR Chemical Shift prediction procedure using B3LYP/cc-pVDZ and empirically derived systematic error correction terms: A Computational Small Molecule Structure Elucidation Method. Dongyue Xin, C. Avery Sader, Om Choudhary, Paul-James Jones, Klaus Wagner, Christofer S. Tautermann, Zheng Yang, Carl A. Busacca, Reggie Saraceno, Keith R. Fandrick, Nina C. Gonnella, Keith Horspool, Gordon Hansen, and Chris H. Senanayake J. Org. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.joc.7b00321 • Publication Date (Web): 11 Apr 2017 Downloaded from http://pubs.acs.org on April 12, 2017
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
The Journal of Organic Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 11
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Organic Chemistry
Development of a 13C NMR Chemical Shift prediction procedure using B3LYP/cc-pVDZ and empirically derived systematic error correction terms: A Computational Small Molecule Structure Elucidation Method. Dongyue Xin,1 C. Avery Sader,1,2 Om Chaudhary,1,2 Paul-James Jones,3 Klaus Wagner,4 Christofer Tautermann,4 Zheng Yang,3 Carl A. Busacca,2 Reggie Saraceno,1 Keith R. Fandrick,2*‡ Nina C. Gonnella,1*‡ Keith Horspool,1 Gordon Hansen,3 Chris H. Senanayake2 1
Material and Analytical Sciences, Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut, USA. 2Chemical Development, Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut, USA. 3Information Technology, Boehringer Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut, USA. 4Boehringer Ingelheim Pharma GmbH & Co.KG, Germany. 3 Deceased (14 August 2015).
Supporting Information Placeholder ABSTRACT: An accurate and efficient procedure was developed for performing 13C NMR chemical shift calculations employing density functional theory with the gauge invariant atomic orbitals (DFT-GIAO). Benchmarking analysis was carried out, incorporating several density functionals and basis sets commonly used for prediction of 13C NMR chemical shifts, from which the B3LYP/cc-pVDZ level of theory was found to provide accurate results at low computational cost. Statistical analyses from a large data set of 13C NMR chemical shifts in DMSO are presented with TMS as the calculated reference and with empirical scaling parameters obtained from a linear regression analysis. Systematic errors were observed locally for key functional groups and carbon types, and correction factors were determined. The application of this process and associated correction factors enabled assignment of the correct structures of therapeutically relevant compounds in cases where experimental data yielded inconclusive or ambiguous results. Overall, the use of B3LYP/cc-pVDZ with linear scaling and correction terms affords a powerful and efficient tool for structure elucidation.
Introduction Structure elucidation, a process that often combines synthetic and spectroscopic methods to prove a proposed structure, plays a critical role in organic chemistry, medicinal chemistry and pharmaceutical development. The combination of high level NMR and Mass spectrometry (analytical) data may be inconclusive at times, leading to a set of proposed structures that require alternative time-consuming strategies for unambiguous structure determination which include systematic syntheses of the various structural possibilities. Such challenges are frequently encountered within the pharmaceutical industry where extensive resources are spent on both spectroscopic and synthetic efforts to derive the structure of an unknown analyte necessary for establishing active pharmaceutical ingredient (API) integrity and drug safety.
NMR spectroscopy, along with mass spectrometry and single crystal X-ray, are powerful technologies used to determine molecular structure. An advantage that NMR spectroscopy has over mass spectrometry is its ability to trace full atomic connectivity in a molecule. NMR spectroscopy also has an advantage over X-ray crystallography in cases where compounds are non-crystalline, only available in microgram quantities, are unstable transient species or contain significant impurities. Advances in NMR technology that include gradient solvent suppression and micro cryoprobe technology have expanded the limits of structure elucidation, enabling such structures to be solved. Nonetheless, challenging cases where results from NMR spectra may be ambiguous or inconclusive are still a frequent occurrence in pharmaceutical research and development.1,2 To address the current challenges in structure elucidation, a practical and robust process is needed that can reliably differentiate between multiple structural possibilities derived from NMR data. The ability to narrow the structural possibilities to a likely candidate and accordingly direct synthetic and analytical efforts on a single structure would result in major cost savings in both time and materials while also accelerating the development of a compound. This is particularly important when dealing with structure elucidation of isolated degradants, impurities, reaction byproducts and metabolites. Often such compounds have unstable or unusual structures or are isolated in low quantities, making structure elucidation by conventional procedures highly challenging.1 The assignment of NMR chemical shifts with quantum mechanical methods has been highly successful in addressing a variety of complex structural uncertainties where experimental data is inconclusive. Examples include determination of relative stereochemistry 3,4, and definitive assignment of regioisomers.5 Tantillo has summarized the technical considerations of employing ab initio and density functional theory calculations for NMR chemical shift prediction,6 while Hoye has developed a more automated approach for 1H and 13C chemical shift prediction that is targeted at the experimental chemist who is not an expert with computational methods.7 Benchmarking studies for the prediction of chemical shifts with varying levels of sophistication have been
ACS Paragon Plus Environment
The Journal of Organic Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
reported.8-11 However, single conformer datasets were exclusively used for accuracy assessment. As far as we are aware, an extensive benchmarking analysis with complex flexible compounds, is currently lacking despite the application of this methodology to increasingly elaborate scaffolds such as natural products and drug-like molecules. Here we report a robust and practical protocol for NMR chemical shift prediction and its application to challenging structural elucidation problems.12 In order to perform the necessary DFT calculations for geometry optimization and shielding constant calculations, parallel processing on a High Performance Computing (HPC) cluster was employed. Because of the time consuming and tedious nature of the process, automation was employed for comprehensive evaluation. This automated application was initially incorporated into the CCG Moe2015 platform.13 A survey of commonly reported density functionals and basis sets used in NMR chemical shift prediction studies was carried out that supported selection of B3LYP/cc-pVDZ as the preferred choice for accuracy and computational efficiency. Corrections regarding both referencing and systematic errors for B3LYP/cc-pVDZ are also presented, which altogether give rise to an accurate approach with a reasonable computational cost. Materials Gaussian14 calculations were performed using a linux workstation interfaced with a high performance grid computing environment based on 64 bit-Scientific Linux 6.6 and composed of a mixture of nodes containing Xeon E5-2670/2.6GHz and E52680/2.8GHz CPUs. The exact distribution of cores in the grid used for any given calculation was determined at runtime by the queuing system based on availability, system load and other factors, and varied from run to run. NMR experiments were performed on a Bruker-Biospin AVIII 600 NMR spectrometer operating at 600.2 MHz for 1H and 150.92 MHz for 13C, nonspinning, using a 1.7mm CP TCI Z gradient probe. 1H and 13C chemical shifts were referenced to DMSO-d6 at 2.5ppm and 39.5ppm, respectively. The NMR measurements, including 1H, 13C, HSQC, HMBC, COSY, and NOESY experiments, were carried out at 300 K in DMSO solutions.15-18 Experimental parameters were as follows: for the 1H spectrum, acquisition time (AQ) = 1.32 s, relaxation delay (RD) = 2.0 s, 90° pulse = 10s, spectral width (SW) = 12376 Hz; for the 13C spectrum, AQ= 0.904 s, RD= 0.5 s, 90° pulse width = 12 us, SW= 36231 Hz; the mixing time for the NOESY spectrum was 0.5 s. Results and Discussion Automation of the Process The procedure used to compute 13C NMR chemical shifts is shown in Figure 1. The structure or a series of structural possibilities of a compound is initially proposed based upon experimental data. A conformational search is then conducted for each structure with the Molecular Operating Environment (MOE 2014.09)13 software using the LowModeMD19 algorithm which uses implicit vibrational analysis to focus an MD trajectory along low-mode vibrations and search for minima along the valleys and troughs on the potential energy surface. RMSD and upper energy limits of 0.75 Å and 5.0 kcal/mol, respectively, were used to generate a computationally tractable number of conformers. After the conformer sets were created, custom scripts were used to extract individual conformers, create subdirectories, and generate input files for optimization with quantum mechanical methods. The entire automated platform including data analysis was first incorporated into the CCG MOE 2015 program.13
Page 2 of 11
Figure 1. Flowchart depicting the methodology for calculating chemical shifts
Geometry optimizations, calculation of analytical frequencies, and calculation of 13C NMR shielding constants were performed at the density functional theory level in Gaussian 09, revision D.0114. The effects of solvation were evaluated implicitly using the conductor-like polarizable continuum model (CPCM)20 with dimethyl sulfoxide (DMSO), chloroform (CDCl3), methanol (MeOH), or acetonitrile (ACN). Frequencies were computed in order to obtain zero-point energies, thermal corrections, and to verify that each conformer is a true potential energy minimum with no imaginary frequencies. Isotropic shielding constants were computed for all optimized conformers that met this stability criterion. The gauge-invariant atomic orbital approach 21,22 was used to compute 13C NMR shielding constants, which were converted to chemical shifts by subtracting them from the shielding constant value of TMS computed at the same level of theory for DMSO: 192.7 used in this study, and also for Chloroform: 192.5, Methanol: 192.7, Acetonitrile: 192.6, or by applying linear scaling factors. Boltzmann-weighting factors for each conformer were determined by using the relative free energies obtained from the frequency calculations. Benchmarking Benchmarking of computational methods used in NMR structure elucidation was performed using a molecule that contains functional groups commonly encountered in pharmaceutical compounds. Compound 1 was selected for this purpose.23,24 The functionals tested were B3LYP25,26, ωB97XD27, M06-2X28, WC0429, and B3LYP-D3 with Becke-Johnson damping30. The Pople-type basis sets 6-31+G(d,p)31,32 and 6-311++G(2d,p), the Ahlrichs-type basis set def2-TZVP33, the correlation consistent cc-pVDZ basis set of Dunning34, and the polarization consistent pcS-2 basis set of Jensen35 were examined with each of the density functionals. The WC04 functional and the pcS family of basis sets were designed to improve accuracy in computing magnetic shielding constants and are not appropriate for calculation of free energies; therefore, single-point GIAO calculations employing these methods were performed using optimized geometries and frequency corrections at the B3LYP/6-311++G(2d,p) level. The accuracy of each functional/basis set combination was compared with root mean square deviation (RMSD). Three-dimensional structures were rendered with CYLview.36 The lowest energy conformer and numbered chemical structure of compound 1 are shown in Figure 2 along with plots that show
ACS Paragon Plus Environment
Page 3 of 11
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Organic Chemistry
the difference between computed and experimental 13C NMR chemical shifts. The global minimum conformer shows a stabilizing intermolecular hydrogen bond of 1.66 Å between the hydrogen of the alcohol group on C12 and the carbonyl oxygen on C24, resulting in a pseudo nine-membered ring with a boat-boat conformation.
Figure 2. Top: Two perspectives of the lowest energy conformer of compound 1 at the B3LYP/cc–pVDZ level of theory with CPCM solvation in dimethyl sulfoxide. Bottom: The chemical structure of 1 with the numbering scheme used herein.
The performance of the computational methods for the prediction of 13C NMR chemical shifts in compound 1 is summarized in Table 1. RMSD values are displayed in ppm. A single set of 25 conformers was used as input; however, convergence failures or the presence of imaginary frequencies prevented the use of all 25 conformers with some functional/basis set combinations. The difference between the best performing method (ωB97XD/ccpVDZ; RMSD = 3.20 ppm) and the worst performing method (M06-2X/pcS-2//B3LYP/6-311++G(2d,p); RMSD = 24.8 ppm) is more than 20 ppm, hence the choice of calculation method is critical to produce acceptable accuracy. Notably, ωB97XD performed better than the other density functionals for sulfone and trifluoromethyl groups. The use of B3LYP-D3 resulted in marginal improvement in the RMSD values with respect to standard B3LYP and no significant qualitative changes in the optimized geometries or difference plots were observed. Our study found that with B3LYP/cc-pVDZ comparable accuracy with other methods examined was achieved at low computational cost. Table 1. Root mean square deviations (ppm) of 13C chemical shift calculations at various levels of theory. basis set functional B3LYP B3LYP-D3 ωB97XD M06-2X WC04b
ccpVDZ
631+G(d,p)
6311++G(2d,p)
def2TZVP
pcS2a
4.06 3.60 3.20 14.02 3.24
4.00 3.86 3.28 11.36 4.61
8.01 7.78 5.99 3.41
8.83 8.71 3.83
15.2 15.2 15.1 24.8 6.10
(CSGT) schemes37, hence this study parallels our basis set analysis findings for 13C NMR predictions. TMS vs. Linear Regression Comparison was made between TMS and the Linear Regression method6 relative to the chemical shift prediction accuracy. Figure 3 shows a pair of histograms that illustrates the distribution of differences between the experimental and calculated chemical shifts referenced to TMS and with linear scaling, determined for a set of 51 organic molecules (766 chemical shifts). Because systematic errors introduced by third row and higher elements in the periodic table can adversely affect linear correlations derived mainly from, and intended for, compounds lacking these heavier elements, known systematic errors were excluded from the histograms in Figure 3. Specifically, nine chemical shifts of halogensubstituted carbons, 17 chemical shifts of carbons bonded to sulfones, ten chemical shifts of carbons bonded to sulfur in aminothiazole heterocycles, and three chemical shifts of carbons bonded to phosphorous were excluded from the analysis (see systematic errors section 4 below). When TMS is used as a reference (Figure 3, top), the distribution is approximately normal with a bias toward negative differences. That is, when a large amount of data is considered, B3LYP/cc-pVDZ on average overestimates the chemical shift with respect to experiment by a mean of -1.5 ppm, which gives reliable proof of a systematic deviation at this level. For ease of analysis, the ranges of chemical shift differences were separated into zones based on standard deviations from the mean of a Gaussian distribution. With a standard deviation of 2.7 ppm, this data shows that 95% of all calculated chemical shifts are expected to be within the range of -6.9 to 3.9 ppm of the correct assignment. About 5% of computed chemical shifts differ from the corresponding experimental shift in the ranges of -9.6 to -6.9 or +3.9 to +6.6. Less than 0.3 % of computed chemical shifts will result in a difference of less than -9.6 ppm or greater than 6.6 ppm. The statistical analysis gives a foundational basis for an acceptable fit to experimental data for the B3LYP/cc-pVDZ method.
a,b
Single point GIAO calculations using B3LYP/6-311++G(2d,p) optimized geometries and frequencies.
Based on our assessment we found that use of the B3LYP/ccpVDZ in 13C chemical shift prediction produced a reliable method balancing accuracy with computational time. Toomsalu and Burk have reported that the aug-cc-pVDZ basis set produces accurate results for 13C chemical shifts with a preference for use of PBE1PBE method and the continuous set of gauge transformation
ACS Paragon Plus Environment
The Journal of Organic Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Figure 3. Histogram of the differences observed between experimental and calculated 13C chemical shifts for a set of 51 organic molecules at the B3LYP/cc-pVDZ level of theory in DMSO. Top graph (δexpt – δcalc; TMS reference); Bottom graph (δ = (intercept isotropic magnetic shielding)/slope; Linear Scaled reference).
The distribution of errors after linear scaling (Figure 3, bottom) is Gaussian which by necessity has a mean of 0.0 ppm. Our data shows that 95% of differences between experimental and computed chemical shifts are expected to be within ±4.8 ppm. This value will play a prominent role in distinguishing the structure of a compound when more than one isomer is possible from available experimental data. A linear fit for calculated 13C isotropic shielding constants plotted against corresponding experimental chemical shifts for our set of 51 small to medium-sized organic molecules is shown in Figure 4. A correlation coefficient greater than 0.995 is necessary to qualify a method as reliable; therefore, the tight correlation shown by the R2 value of 0.9979 indicates almost no random error associated with B3LYP/cc-pVDZ, where carbon atoms that introduce systematic errors are excluded. The deviation of the slope from unity is another indicator of systematic error in the method. Typically this slope should be within the range of 0.95-1.05 for a reliable method, which is in fact observed here.6,11
Page 4 of 11
Linear scaling factors for B3LYP/cc-pVDZ in DMSO obtained from this analysis are 188.57 with a slope of -0.9759. Linear scaling factors for B3LYP/cc-pVDZ in chloroform (189.29/-0.9710), acetonitrile (189.66/-0.9705) and methanol (190.59/-0.9766) were also obtained (data not shown). Either TMS or linear regression may be used as chemical shift references with comparable overall conclusions. Throughout the text, our method of choice for the absolute assignment of a given compound is the TMS reference, unless otherwise stated. Since linear regression consistently performs better than TMS, the corresponding linear regression data is also provided in the supporting information. Systematic Errors Plots of the differences between experimental and computed 13C chemical shifts for compound 1 at the B3LYP/cc-pVDZ level of theory are shown in Figure 5. The computed isotropic shielding constants were converted to chemical shifts with the TMS reference. With the TMS reference, it is clear that sulfone carbons C2 and C10 are predicted to be more deshielded relative to experimental data due to heavy atom effects of sulfur mentioned previously. As illustrated in Figure 5, computed chemical shifts of C13 in CF3 display similar deviations from experimental data at the B3LYP/cc-pVDZ level. The predicted chemical shifts of carbons C2, C10 and C13, along with C11, C12, C14, and C15 in the flexible alkyl chain show relatively worse agreement than the rest of the molecule. It is apparent with the pharmaceutical representative molecule that the deviations from prediction of the chemical shifts are localized alpha to certain functionalities (sulfone in particular). These deviations are due to limitations of theory, solvation modeling or the combination of the two. However, if these deviations are systematic an appropriate correction factor could be employed to compensate for these deviations while also preserving the balance between computational resources and accuracy as described previously.
Figure 5. Plots of the differences between the calculated and experimental 13C NMR chemical shifts of 1. Shielding constants were computed at the B3LYP/cc-pVDZ level of theory and converted to chemical shifts referenced by TMS. Values of RMSD are given in ppm.
Figure 4. Correlation plot of experimental 13C chemical shifts and calculated shielding constants of a set of 51 organic molecules at B3LYP/cc-pVDZ in DMSO.
In examining data from a set of 69 pharmaceutically relevant compounds with B3LYP/cc-pVDZ, consistent errors are observed for several functional groups and atom types. We have also examined data for halogen bearing aromatics and found correction terms for halogen bearing carbons. While it is known that halogens introduce spin orbital effects into the 13C chemical shift calculations38, no systematic investigation of these errors with regards to complex structures have been reported. We obtained correction terms for C-F, C-Cl, and C-Br with the computationally
ACS Paragon Plus Environment
Page 5 of 11
efficient B3LYP/cc-pVDZ and for C-I with B3LYP/lanl2dz using 137 compounds and >137 chemical shifts. As shown in Figure 6 the errors are clustered within a narrow range of values and thus are amenable to an appropriate correction term. C*-F
C*-Cl
C*-Br
C*-I
10 0
ment can be realized in applying the correction terms. The postcomputational application of the derived correction factors for these chemical shifts greatly enhances the accuracy of the prediction without incurring additional computational cost. Furthermore, the application of these correction factors allows all the predicted chemical shifts to be within 5 ppm of experimental values thus these correction factors will allow a higher confidence in the discernment of structural possibilities where the correct structure is not known.
-10 Δδ (ppm)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Organic Chemistry
-20 -30 -40 -50 -60 -70
Figure 6. Graph of chemical shift deviations (ppm) for halogenated aromatic compounds. Correction terms were derived from the mean calculated differences [
∑𝑛𝑛 𝑖𝑖=0(𝑒𝑒𝑒𝑒𝑒𝑒−𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐) 𝑛𝑛
]
The different functional groups and atom types that showed consistent deviation from experimental data included aromatic carboxylic acids, carbon shifts alpha to a sulfone groups, carbon shifts alpha to sulfur in thienopyrimidines, tertiary carbons alpha to phosphine oxides, CF3, aminothiazoles and quaternary carbons. Although these errors result from a complex set of interactions (solvation, limitation of theory, etc), the inherent errors appear systematic and thus potentially compensated by a suitable postcomputational addition factor. A summary of correction terms is listed in Table 2.
Table 2. Correction Terms for Selected Functional Groups Functionality
Mean Correction + 9.9 ppm + 20.2 ppm - 4.0 ppm + 8.9 ppm + 6.8 ppm + 9.3 ppm + 8.3 ppm
Number of Compounds >50 >50 38 25 6 12 15
Maximum Deviation without Correction -13.5 ppm -24.7 ppm 6.8 ppm -16.2 ppm -7.5 ppm -15.0 ppm -12.2 ppm
Cl Br COOH Sulfone Groups CF3 Aminothiazoleb Quaternary Thieno[3,2+7.7 ppm 10 -8.1 ppm d]pyrimidine C4c Thieno[3,2+4.9 ppm 10 - 5.3 ppm d]pyrimidine C1c t-Butylphosphine +12.2 ppm 8 -12.6 ppm Oxided a All data collected at the B3LYP/cc-PVDZ level of theory using the TMS reference. b)
R
c) N
R R
R R * S
d)
N C1
N
C4 S
O P * R R
R
For the test compound 1 (Figure 7, blue bars) several functionalities are apparent that were shown to be systematically deficient in the predicted NMR chemical shifts (sulfones, trifluormethyl and quaternary carbon). When experimentally determined systematic error corrections were applied for B3LYP/cc-PVDZ, the RMSD becomes 2.29 ppm, hence significant and reproducible improve-
Figure 7. Plots of the differences between the calculated and experimental 13C NMR chemical shifts of 1 before (top) and after (bottom) applying systematic error corrections. Shielding constants were computed at the B3LYP/cc-pVDZ level of theory and converted to chemical shifts referenced to TMS. Values of RMSD are given in ppm.
Demonstration of an applied systematic error correction term is evident from our studies on monobromophakellin39, a pyrroleimidazole alkaloid (PIA) natural product (Figure 8, 2a). The 13C NMR chemical shift calculation revealed a significant deviation (22.4 ppm) for 2a at C4 where Br is attached. After applying + 20.2 ppm correction, the deviation at C4 was within 5 ppm of the experimental value. The 13C NMR prediction calculations for monobromophakellin also showed differences between experimental and computed 13C chemical shifts for 2a and its protonated species 2b. A significant deviation of -12.2 ppm was observed at C11 for 2a while deviations were within three standard deviations from the mean for 2b (Figure 8b). These calculations support a protonated monobromophakellin which shows better agreement with experimental data than monobromophakellin (Figure 8c). The results and conclusions were consistent with our previous study using 15N chemical shift prediction where the protonated structure of monobromophakellin was found to agree with experimental data.12
ACS Paragon Plus Environment
The Journal of Organic Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The data, plotted in Figures 8a-c, illustrate the correction term for Br satisfied calculations for 2b (Figure 8b) as well as 2a (Figure 8c). A red box over a position number serves as a literal “red flag” by indicating a difference that is greater than three standard deviations away from the mean of -1.5 ppm, an observation that has a probability of 0.3 %. A yellow box indicates that the observed difference falls between two and three standard deviations away from the mean, which has a probability of ~5 %.
5
4 Br
H 12 N
14 N
H2N
11
N 10 9
Br
3 2
N 6 O
H N
N
N H
N
Page 6 of 11
TIC-10 (TRAIL inducing compound 10) TRAIL (TNF related apoptosis inducing ligand) is a cytokine that kills cancer cells but shows no toxicity against normal cells. Active pharmaceutical ingredient TIC-10 (TRAIL inducing compound 10) (compound 3a), is an anticancer drug first reported in a 1973 patent.40 The structure of TIC10 3a had been mis-assigned as 3b in the initial patent. Subsequent synthesis of the originally reported structure by an alternate route showed compound 3b did not possess the reported biological activity of inducing TRIAL expression. This finding initiated structure elucidation of the active material using NMR spectroscopy and single crystal X-ray revealing 3a as the correct structure.41 A third isomer of TIC10 3c was also reported that proved unable to induce TRIAL expression. Numbered chemical structures for 3a, 3b, and 3c are given in Figure 9.
H2N O
8
2a monobromophakellin
2b protonated monobromophakellin
Figure 9. Left: The lowest energy conformer of the active isomer of Tic10 (3a) and two other possible isomers (3b-c) at the B3LYP/cc–pVDZ level of theory with CPCM solvation in dimethyl sulfoxide. Right: The numbered chemical structure of each isomer.
Figure 8. Plots of the differences between the calculated and experimental 13C NMR chemical shifts of 2b before (a) and after (b) applying systematic error corrections for Br (+ 20.2 ppm) and 2a after applying Br corrections (c). Shielding constants were computed at the B3LYP/cc-pVDZ level of theory and converted to chemical shifts referenced to TMS. Values of RMSD are given in ppm.
The TIC-10 structure 3a and two regioisomers were initially analyzed via mass spectrometry, but full characterization had not been carried out leading to compromised structure evaluation.41 Because structural mis-assignment can incur significant cost with respect to patent issues, biological activity and pharmacophore development, determining the correct structure is critical in both the discovery and development process. Topological assignment of the respective carbon atoms in the given structure and structural possibilities is paramount for accurate discernment between experimental and calculated chemical shifts. The utilization of all available chemical and analytical evidence is required for full structural characterization. For Tic10 (Figure 9, compound 3a), the three structural possibilities arise from two main precursors: a piperidine carboxylate and two regioisomeric dihydro amino-imidazole systems. Based on the precursors, the mapping of the individual subunits (Figure 10, color coded) for the three possibilities is evident. Given these three
ACS Paragon Plus Environment
Page 7 of 11
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Organic Chemistry
structure possibilities, NMR data becomes imperative in tracing the connectivity for each of the isomers.
N
O
NH NH
+
RO2C
+
N
N NH2
N
Pathway C Pathway A Pathway B O N
O N
N 3a
O N
N N
N 3b
N
N N
N
N
3c
Figure 10. Chemical pathways for the three structural possibilities of Tic10 3a-3c along with color-coded substructure relationships. Although both NMR and X-ray data had been used to determine 3a as the active compound, as well as structures of 3b and 3c, the similarities of these regioisomers made this an interesting application to evaluate if the correct regioisomer could be determined given the three possibilities with only one set of 13C NMR chemical shifts. Therefore DFT calculations at B3LYP/cc-pVDZ level of theory were performed for all three isomers and calculated chemical shifts were compared with experimental data from the biologically active material. The experimental chemical shifts and the calculated chemical shifts of 3a, 3b and 3c are listed in Table 3. The results, graphed in Figure 11, show good agreement between experimental and prediction for compound 3a (< 5 ppm) with some larger deviations (> 5 ppm) clearly evident for 3b and 3c. Table 3. Comparison of experimental 13C chemical shifts with calculated chemical shifts for Tic10 isomers 3a-c Position δexpt δa,calc δb,calc δc,calc 1 160.8 159.3 155.9 156.2 2 46.3 48.9 47.6 42.7 3 50.0 53.1 52.3 47.3 4 152.1 151.4 150.8 154.2 5 147.2 148.2 149.7 161.7 6 25.8 30.6 29.5 36.3 7 48.3 51.0 51.3 52.9 8 48.7 52.7 52.3 52.5 9 99.2 103.2 107.5 111.6 10 61.3 65.2 64.9 65.8 11 138.2 141.5 141.4 141.9 12 128.7 127.6 127.65 127.7 13 128.2 128 128 127.9 14 127.0 126.5 126.6 126.4 15 42.2 43.5 47.7 49.5 16 134.9 139.1 136.5 136.4 17 124.8 128.9 122.5 128.2 18 125.6 125.4 125.7 125.6 19 126.4 126.7 126.4 127.4 20 129.8 129.4 130 130.1 21 135.0 139.2 136.7 139 22 18.7 21.6 21.5 21.4 RMSD 2.80 3.14 5.57
Figure 11. Plots of the differences between the computed chemical shifts of possible isomers of Tic10 and experimental chemical shifts of active isomer 3a. Computed at the B3LYP/cc–pVDZ level of theory with CPCM solvation in dimethyl sulfoxide with TMS reference. Difference plots for each isomer shown in Figure 11 were generated using the experimental chemical shifts shown in Table 3. A cursory visual inspection is sufficient to disqualify structure 3c as the observed product due to the differences of -14.5 ppm and 12.4 ppm at C5 and C9, respectively. Between 3a and 3b, the difference of -4.0 ppm and -8.3 ppm at C9 provides positive evidence in support of 3a, although within the margin of error for computing chemical shifts with density functional theory. Looking at the plots in the context of our statistical analysis (Figure 3) supports the significant differentiation between 3a and 3b. In the plot of 3b, there are differences at C1 and the alpha position C9 that are larger than two standard deviations from the mean. With 22 13C chemical shifts in Tic10, 9 % of the differences in 3b occur within the 2-3σ range (Figure 3, top), which is almost double the expected probability of 5%. Coupled with the fact that all differences between experimental data and the computed shifts of 3a are less than 2σ, the set of experimental chemical shifts can be reasonably assigned to structure 3a. Nevirapine Hydrolysis Product Nevirapine is a potent inhibitor of HIV-1 reverse transcriptase (RT) which is required for early proviral DNA synthesis42. Because RT is a prime target for antiviral therapy against acquired immunodeficiency syndrome (AIDS), inhibitors of this enzyme were developed for treatment of HIV-1 infection.43 During the development of Nevirapine, a hydrolysis product of the API was observed. The nevirapine hydrolysis product (NHP) required structure elucidation to ensure safety and efficacy in the drug development process. Three separate points of hydrolysis were proposed for Nevirapine (Figure 12) generating structures 4a-4c. Two of the possibilities (4b and 4c) exist as a pair of tautomers between the pyridinone and pyridinol isomers. Both NMR and MS/MS analysis of NHP were performed; however due to the poor solubility of the material in multiple solvents critical information was lost in the
ACS Paragon Plus Environment
The Journal of Organic Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
2D-NMR data. Furthermore, considering the three possibilities 4a-4c, ambiguity arose between MS and NMR data. The MS data showed loss of m/z 44 which supported the presence of a carboxylic acid, making 4a the preferred choice. The 2D NMR COSY data supported the presence of a cyclopropylamine moiety consistent with structures 4b and 4c. Due to this ambiguity between MS and NMR data, a fourth structure was proposed (4d) that arises from a Smiles Rearrangement44 of 4a. Still this new structure (Figure 12 and 13, compound 4d) could not be unequivocally confirmed with NMR and MS data alone. NH2 N
N 4a
O
H N
Smiles Rearrangement
N
N CO2H
CO2H
N NH
4d
HN
N
N
H N
N
Nevirapine
N
O NH
H N
NH Tautomerization N
O
O NH
N OH
4b
HN H N N H
O
N
HN H N
Tautomerization
O 4c
N
Page 8 of 11
Table 4. Comparison of the computed 13C-NMR chemical shifts of 4a-d with experimental shifts for NHP.* δexpt δa,calc δb,calc δc,calc δd,calc Position 1 156.8 145.6 153.9 154.3 155.7 2 119.5 143.4 119.8 128.1 118.7 3 143.6 136.2 142.7 142.6 146.7 4 114.4 126 115.4 109.9 115.1 5 144.2 133.1 144.8 127.8 145.7 6 17.5 19.6 21.1 22.8 20.6 7 169.2 164.4 158.7 164.9 166.5 8 157.6 159.6 159.1 158.9 157.8 9 109.7 129 121.7 109.9 105.5 10 139.8 142.5 144.2 135.7 140.5 11 112.7 122.4 107.1 110.9 112.5 12 151.9 151.1 138.6 152.3 154.3 13 24.3 34.3 28 26.9 27.9 14 6.6 9.9 9.7 9.7 9.9 RMSD 10.78 6.16 5.70 2.37 *TMS reference used in chemical shift calculation.
N
O OH
Figure 12. The structural possibilities and pathways for the proposed structure of the Nevirapine Hydrolysis Product (NHP) and color-coded substructure alignment. To address this problem, computed 13C NMR chemical shifts (TMS reference) of the four potential isomers of NHP (B3LYP/cc-pVDZ) were obtained and comparison was made with the experimental data. The experimental and computed chemical shifts are listed in Table 4. Difference plots (Figure 14) show structures 4a-c all have at least one difference that is larger than three standard deviations from the mean. The plot of 4a exhibits significant chemical shift deviations from experiment for C1, C2, C3, C4, C5, C7, C9, C11 and C13. For compound 4b C7, C9, C11 and C12 showed major deviation from experimental values. Likewise 4c showed outliers from experiment for C2, C4, C5, C7 and C10. As a result, 4a, 4b and 4c could be eliminated as possible candidates with a high level of confidence. For compound 4d however, there was good agreement between computed and experimental values even for C9 which is within expected experimental error. Structure 4d was confirmed via chemical synthesis showing identical NMR spectral comparison.
Figure 13. The lowest energy conformer of NHP (4d) and three other possible isomers (4a-c) at the B3LYP/cc–pVDZ level of theory with CPCM solvation in dimethyl sulfoxide.
ACS Paragon Plus Environment
Page 9 of 11
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Organic Chemistry
Figure 14. Plots of the differences between the computed 13C chemical shifts of 4a-d and the experimental shifts of NHP. Computed at the B3LYP/cc–pVDZ level of theory with CPCM solvation in dimethyl sulfoxide with TMS reference. Compound 4c was also synthesized and characterized by NMR spectroscopy in dimethyl sulfoxide-d6. NMR data showed a preference in the tautomeric equilibrium between the pyridinone (4c) and pyridinol (4c-OH) (Figure 15) that favored the pyridinone 4c. Computations at the B3LYP/cc-pVDZ level of theory predict that 4c is more stable than the aromatic tautomer by 5.8 kcal/mol. Comparison of the computed chemical shifts of 4c and 4c-OH with the available experimental shifts shows a substantial difference in C1-C5 of the ring, favoring the pyridinone (Figure 16a and b). Hence computations at the B3LYP/cc-pVDZ level of theory have been accurate and efficient in determining both regiochemistry and preferred tautomer. It should be noted that the use of the linear regression reference (Figure 16c), instead of TMS (Figure 16a) produced better agreement between calculated and experiment for the correct structure. 14
14
13
13
HN H N 7
6 3 4 5
2
N H
1
∆G =
O
8
N
HN H N 7
6
12
3 9
O
11
4
2
10 5
N
1
8
9
O OH
4c
4c-OH
0.0
5.8
N
12 11
10
Figure 16. Plots of the differences between the computed 13C chemical shifts of 4c and 4c-OH and the experimental shifts of 4c. Computed at the B3LYP/cc–pVDZ level of theory with CPCM solvation in dimethyl sulfoxide with TMS reference (a, b) and linear scaling reference (c, d).
Figure 15. Tautomeric equilibrium of 4c. Relative Gibbs free energies are in kcal/mol. Conclusion We have developed a practical approach to high level NMR chemical shift prediction with an emphasis on pharmaceutically relevant structures. Use of B3LYP/cc-pVDZ afforded a robust and accurate method for chemical shift prediction with low computational cost. For a given functional, the cc-pVDZ basis set is recommended as an acceptable balance between accuracy and computational cost. Because B3LYP remains the most commonly known and employed functional for computational experts and novices alike and it was our method of choice in this study. An evaluation of the TMS reference revealed an average -1.5 ppm deviation from experimental chemical shifts in DMSO. Average TMS and LR reference terms for four solvents, and systematic error corrections for ten functional groups were identified and calculated. Although the method described herein has proven to be robust and broadly applicable for a wide range of organic compounds, we do acknowledge that chemical shift prediction is primarily a guide that can accelerate the structure elucidation process but only when used in conjunction with a totality of chemical and experimental data. With these considerations, the combined use of our preferred method with associated LR or TMS reference and
ACS Paragon Plus Environment
The Journal of Organic Chemistry
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
correction terms provides a powerful and reliable NMR prediction tool that has been automated for the practicing organic chemist.
ASSOCIATED CONTENT Supporting Information The Supporting Information is available free of charge on the ACS Publications website.
Gaussian input files, Cartesian coordinates, energy, isotropic shielding constants and chemical shifts.
AUTHOR INFORMATION Corresponding Authors
[email protected] [email protected] Author Contributions ‡These authors contributed equally.
Notes The authors declare no competing financial interests.
ACKNOWLEDGMENT This paper is dedicated to the memory of Gordon Hansen.
REFERENCES (1) Nicolaou, K. C.; Snyder, S. A. Angew Chem Int Ed Engl 2005, 44, 1012. (2) Gonnella, N. C.; LC-NMR: Expanding the limits of Structure Elucidation, Taylor and Francis: 2013, p 165. (3) Saielli, G.; Nicolaou, K. C.; Ortiz, A.; Zhang, H.; Bagno, A. J Am Chem Soc 2011, 133, 6072. (4) Smith, S. G.; Goodman, J. M. J Am Chem Soc 2010, 132, 12946. (5) Chimichi, S.; Boccalini, M.; Matteucci, A.; Kharlamov, S. V.; Latypov, S. K.; Sinyashin, O. G. Magn Reson Chem 2010, 48, 607. (6) Lodewyk, M. W.; Siebert, M. R.; Tantillo, D. J. Chem Rev 2012, 112, 1839. (7) Willoughby, P. H.; Jansma, M. J.; Hoye, T. R. Nature Protocols 2014, 9, 643. (8) Flaig, D.; Maurer, M.; Hanni, M.; Braunger, K.; Kick, L.; Thubauville, M.; Ochsenfeld, C. J Chem Theory Comput 2014, 10, 572. (9) Teale, A. M.; Lutnæs, O. B.; Helgaker, T.; Tozer, D. J.; Gauss, J. J. Chem. Phys. 2013, 138, 024111. (10) Jain, R.; Bally, T.; Rablen, P. R. J. Org. Chem. 2009, 74, 4017. (11) Pierens, G. K. J Comput Chem 2014, 35, 1388. (12) Xin, D.; Sader, C. A.; Fischer, U.; Wagner, K.; Jones, P. J.; Xing, M.; Fandrick, K. R.; Gonnella, N. C. Org Biomol Chem 2017, 15, 928. (13) Molecular Operating Environment (MOE) (2015) 2014.09; Chemical Computing Group Inc., 1010 Sherbooke St. West, Suite#910, Montreal, QC, Canada, H3A 2R7 (14) Gaussian 09, Revision D.01, Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A., Jr.; Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin, K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.; Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega, N.; Millam, J. M.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.;
Page 10 of 11
Ochterski, J. W.; Martin, R. L.; Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, Ö.; Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J.; Gaussian Inc., 2009. (15) Aue, W. P.; Bartholdi, E.; Ernst, R. R. J. Chem. Phys. 1976, 64, 2229. (16) Jeener, J.; Meier, B. H.; Bachmann, P.; Ernst, R. R. J. Chem. Phys. 1979, 71, 4546. (17) Bax, A.; Davis, D. G. J. Magn. Res. 1985, 63, 207. (18) Bax, A.; Summers, M. F. J. Am. Chem. Soc. 1986, 108, 2093. (19) Labute, P. J Chem Inf Model 2010, 50, 792. (20) Barone, V.; Cossi, M. J. Phys. Chem. A 1998, 102, 1995. (21) Ditchfield, R. Molecular Physics 1974, 27, 789. (22) Kupka, T.; Ruscic, B.; Botto, R. E. J. Phys. Chem. A 2002, 106, 10396. (23) Reeves, J. T.; Fandrick, D. R.; Tan, Z.; Song, J. J.; Rodriguez, S.; Qu, B.; Kim, S.; Niemeier, O.; Li, Z.; Byrne, D.; Campbell, S.; Chitroda, A.; DeCroos, P.; Fachinger, T.; Fuchs, V.; Gonnella, N. C.; Grinberg, N.; Haddad, N.; Jäger, B.; Lee, H.; Lorenz, J. C.; Ma, S.; Narayanan, B. A.; Nummy, L. J.; Premasiri, A.; Roschangar, F.; Sarvestani, M.; Shen, S.; Spinelli, E.; Sun, X.; Varsolona, R. J.; Yee, N.; Brenner, M.; Senanayake, C. H. J. Org. Chem. 2013, 78, 3616. (24) Betageri, R.; Bosanac, T.; Burke, M. J.; Harcken, C.; Kim, S.; Kuzmich, D.; Lee, T. W. H.; Li, Z.; Liu, P.; Lord, J.; Razavi, H.; Reeves, J. T.; Thomson, D. PCT. Int. Appl. WO/2009/149139, 2009. (25) Becke, A. D. J. Chem. Phys. 1993, 98, 5648. (26) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785. (27) Chai, J. D.; Head-Gordon, M. Phys Chem Chem Phys 2008, 10, 6615. (28) Zhao, Y.; Truhlar, D. G. Theor. Chem. Account 2008, 120, 215. (29) Wiitala, K. W.; Hoye, T. R.; Cramer, C. J. J Chem Theory Comput 2006, 2, 1085. (30) Grimme, S.; Ehrlich, S.; Goerigk, L. J Comput Chem 2011, 32, 1456. (31) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56, 2257. (32) Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; Gordon, M. S.; DeFrees, D. J.; Pople, J. A. J. Chem. Phys. 1982, 77, 3654. (33) Weigend, F.; Ahlrichs, R. Phys Chem Chem Phys 2005, 7, 3297. (34) Dunning, J. T. H. J. Chem. Phys. 1989, 90, 1007. (35) Jensen, F. J Chem Theory Comput 2008, 4, 719. (36) Legault, C. Y.; 1.0b ed.; Université de Sherbrooke: 2009. (37) Tormena, C. F.; Silva, G. V. J. d. Chem. Phys. Lett. 2004, 398, 466. (38) Kaupp, M.; Malkina, O. L.; Malkin, V. G. Chem. Phys. Lett. 1997, 265, 55. (39) Meyer, S. W.; Kock, M. J Nat Prod 2008, 71, 1524. (40) Staehle, H. D.; Koeppe, H. D.; Kummer, W. D.; DE Patent 2150062 A1 Apr 12, 1973.. (41) Jacob, N. T.; Lockner, J. W.; Kravchenko, V. V.; Janda, K. D. Angew Chem Int Ed Engl 2014, 53, 6628. (42) Gilboa, E.; Mitra, S. W.; Goff, S.; Baltimore, D. Cell 1979, 18, 93. (43) Hargrave, K. D.; Proudfoot, J. R.; Grozinger, K. G.; Cullen, E.; Kapadia, S. R.; Patel, U. R.; Fuchs, V. U.; Mauldin, S. C.; Vitous, J.; Behnke, M. L.; Klunder, J.M.; Pal, K.; Skiles, J.W.; McNeil, D.W.; Rose, J.M.; Chow, G.C.; Skoog, M.T.; Wu, J.C.; Schmidt, G.; Engel, W.W.; Eberlein, W.G.; Saboe, T.D.; Campbell, S.J.; Rosenthaland, A.S.; Adams, , J J Med Chem 1991, 34, 2231. (44) Levy, A. A.; Rains, H. C.; Smiles, S. J. Chem. Soc. 1931, 3264.
ACS Paragon Plus Environment
Page 11 of 11
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The Journal of Organic Chemistry
. O HN
N Nevirapine
+ H2O, H , ∆
N
?
N
XX X H N
N
O NH
HN H N
NH
O
N H
O
O
NH2
N
N
H N
N
N
N
CO2H
ACS Paragon Plus Environment
NH
CO2H
N