Computer-assisted ¹H NMR analysis of the anti-tuberculosis drug lead ecumicin

Introduction

Ecumicin (1) is a drug lead compound that was recently discovered as part of a high-throughput screening campaign for new narrow spectrum anti-tuberculosis (TB) agents from an extensive actinomycete extract library.^[1] As a macrocyclic tridecapeptide, 1 exerts its potent bactericidal activity by targeting the mycobacterial hexameric chaperon protein, ClpC1.^[2] This represents a new anti-TB drug target that is not shared by any current TB drug and explains why 1 is effective in vitro, and potentially in vivo, for the treatment of even drug-resistant TB (Fig. 1).^[2]

Figure 1.

The structure of ecumicin (1).

With a molecular formula of C₈₃H₁₃₄N₁₄O₁₇, the 1D ¹H NMR spectrum of ecumicin is rather complex. Even when using a protic NMR solvent, methanol-d₄, and acquiring the spectrum with an ultra-high-field instrumentation (900 MHz), dispersion remains intrinsically limited, preventing full interpretation of the complex resonance patterns. As a result, even careful visual analysis restricted the reporting of 13 proton signals as multiplets and three as broad singlets (Fig. 2).^[1]

Figure 2.

Result of the ¹H iterative full spin analysis (HiFSA) of the 900 MHz ¹H spectrum of ecumicin (1) (experimental vs. simulated spectrum).

In the present work, complete ¹H NMR assignments of 1 were achieved by means of ¹H iterative full spin analysis (HiFSA).^[3] Using the iterator of the PERCH NMR software (PERCH Solutions Inc., Kuopio, Finland) tool, this led to the complete and accurate determination of all chemical shifts (δ_H), ¹H,¹H spin–spin coupling constants (J_HH), and experimental line widths, thus greatly complementing conventional NMR data examination.^[4] This uniquely comprehensive ¹H spin information can serve as a basis for all subsequent NMR-based studies of 1 and its analogues. Importantly, the data can be adapted across all available NMR field strengths. In addition, this information facilitates quantitative NMR analyses of these compounds, e.g. for pre-clinical drug development studies.^[3,5]

Results and discussion

The isolation and structure elucidation of 1 (Fig. 1) has recently been reported.^[1] Briefly, ecumicin (1) was isolated from the mycelial methanolic extract of Nonomuraea sp. MJM5123 via a fractionation scheme of three sequential steps, guided by assays monitoring the activity against Mycobacterium tuberculosis and toxicity against mammalian cells. Extensive analysis of NMR experiments resulted in two plausible planar structures, and the ambiguity was solved by LC-MS². The absolute configuration of 1 was determined by Marfey’s analysis, and the structure was ultimately confirmed by X-ray crystallography.^[1]

The protic solvent, methanol-d₄, was chosen in order to simplify the heavily overlapped ¹H NMR spectrum of 1. Although the use of an aprotic NMR solvent such as DMSO-d₆ would provide information about the exchangeable protons and, thus, the amino acid sequence, the combination of a protic NMR solvent and an HMBC experiment is equally capable of uncovering the sequence information. In addition, in aprotic NMR solvents, the exchangeable amide protons tend to give rise to broad signals, which may overlap with side chain ¹H signals and complicate the NMR spectrum further; in contrast, aprotic solvents usually yield sharper ¹H NMR signals.

It was likely that a molecule’s conformation in solution differs from that in its crystalline form. To explore potential differences, molecular modeling calculations were performed. However, the resulting structure was similar to the X-ray structure of 1 (Supporting Information), suggesting that the conformation of ecumicin is relatively rigid, with the main flexibility involving the three amino acid tail. Although the ACA graphical user interface of the PERCH NMR software has been successfully used to analyze the ¹H NMR spectra of many molecules in a semi-automated or fully automated manner, including the nonapeptide atosiban,^[6] it was not possible to analyze the NMR spectrum of 1 in an unattended process. This may be in part due to the large number of nuclei that must be simultaneously assessed, which increases dramatically the number of potential solutions to be evaluated by the iteration software. Moreover, the presence of multiple structural moieties with identical coupling patterns and similar chemical shift distributions, such as five l-V residues, augments the chances of misassignment under a fully automatic regime. While the predicted chemical shifts were suitable starting values (Table S1), and more suitable than reported values of free amino acids,^[7] manual adjustments of the starting parameters were necessary and required careful examination of the 1D and 2D NMR spectra of 1.^[1] The severe signal overlap in the aliphatic region hindered visual analysis the most. Nevertheless, HiFSA was able to deconvolute these signals, thus allowing for a thorough interpretation of all overlapped resonances. As a result, accurate δ_H and J_HH values were obtained for all protons (Tables 1 and 2), including those which previously were reported as overlapped ‘multiplets’.

Table 1.

Final HiFSA-based ¹H chemical shifts (δ_H, in ppm) and multiplicities of ecumicin (1)

Amino acid	Position	δH (ppm)	Multiplicity	Amino acid	Position	δH (ppm)	Multiplicity
N,N-Me2-l-V	2	2.672	d	l-V3	2	4.594	d
	3a	2.042	dqq		3a	2.027	dqq
	4	0.847	d		4	0.915	d
	5	0.978	d		5	0.859	d
	6,7 (N,N-Me2)	2.313	s
				N-Me-l-V	2	3.065	d
l-V1	2	4.669	d		3a	2.583	dqq
	3a	2.101	dqq		4	1.094	d
	4	0.992	d		5	0.976	d
	5	1.055	d		6 (N-Me)	3.140	s
N-Me-l-allo-I	2	4.918	d	N-Me-4-OMe-l-W	2	4.102	d
	3a	1.957	dddq		3a	3.545	dd
	4aa	0.989	ddq		3b	3.694	dd
	4ba	1.246	ddq		5b	6.698	d
	5	0.754	dd		7c	6.918	ddd
	6	0.744	d		8	6.984	dd
	7 (N-Me)	3.234	s		9	6.442	dd
					12 (O-Me)	3.826	s
l-T	2	5.167	d		13 (N-Me)	2.157	s
	3	5.783	dq
	4	1.313	d	l-V4	2	4.529	d
					3a	2.200	dqq
N-Me-l-T	2	5.018	d		4	1.028	d
	3	4.457	dq		5	0.989	d
	4	0.912	d
	5 (N-Me)	3.330	s	Ph-l-threo-S	2	4.854	d
					3	5.344	d
l-V2	2	4.842	d		5 (o)b	7.240	dddd
	3a	2.350	dqq		6 (m)b	7.255	dddd
	4	1.092	d		7 (p)b	7.201	dddd
	5	0.979	d		8 (m)b	7.255	dddd
					9 (o)b	7.240	dddd
N-Me-l-L
	2	5.110	d
	3aa	1.453	ddd	l-V5	2	4.396	d
	3ba	1.239	ddd		3a	1.968	dqq
	4a	0.957	ddqq		4	0.935	d
	5	0.168	d		5	0.919	d
	6	0.331	d
	7 (N-Me)	3.259	s

^aDefined multiplicities are assigned to signals previously reported as multiplets (m).^[1]

^bDefined multiplicities are assigned to signals previously reported as singlets (s) or broad singlets (br s).^[1]

^cHiFSA reveals the signal previously assigned as a doublet of doublets (dd)^[1] to be a ddd.

Table 2.

HiFSA-based ¹H,¹H spin–spin coupling constants (J_HH, in Hz) of ecumicin (1)

Amino acid	Coupling	JHH (Hz)	Amino acid	Coupling	JHH (Hz)
N,N-Me2-l-V	3J (2,3)	9.18	l-V3	3J (2,3)	8.88
	3J (3,4)	6.59		3J (3,4)	6.55
	3J (3,5)	6.62		3J (3,5)	6.78
l-V1	3J (2,3)	8.75	N-Me-l-V	3J (2,3)	7.62
	3J (3,4)	6.79		3J (3,4)	6.50
	3J (3,5)	6.72		3J (3,5)	6.82
N-Me-l-L	3J (2,3)	11.23	N-Me-4-OMe-l-W	3J (2,3a)	11.16
	3J (3,4a)a	1.56		3J (2,3b)	4.71
	3J (3,4b)a	3.32		2J (3a,3b)	−13.73
	3J (3,6)	6.63		5J (5,7)a	0.48
	2J (4a,4b)a	−12.91		4J (7,9)a	0.67
	3J (4a,5)	7.26		3J (8,9)	7.79
	3J (4b,5)a	7.64		3J (7,8)	8.18
l-T	3J (2,3)	2.34	l-V4	3J (2,3)	7.91
	3J (3,4)	6.51		3J (3,4)	6.75
				3J (3,5)	6.81
N-Me-l-T	3J (2,3)	3.73
	3J (3,4)	6.46	Ph-l-threo-S	3J (2,3)	1.90
				4J [5(o),9(o)]a	1.33
l-V2	3J (2,3)	8.92		3J [5(o),6(m)]a	7.65
	3J (3,4)	6.69		3J [8(m),9(o)]a	7.65
	3J (3,5)	7.01		5J [5(o),8(m)]a	0.76
				5J [6(m),9(o)]a	0.76
N-Me-l-L	3J (2,3a)	8.58		4J [5(o),7(p)]a	1.25
	3J (2,3b)	6.59		4J [7(p),9(o)]a	1.25
	2J (3a,3b)a	−13.56		4J [6(m),8(m)]a	1.71
	3J (3a,4)a	5.44		3J [6(m),7(p)]a	7.47
	3J (3b,4)a	7.99		3J [7(p),8(m)]a	7.47
	3J (4,5)	6.55
	3J (4,6)	6.59	l-V5	3J (2,3)	8.90
				3J (3,4)	6.83
				3J (3,5)	6.41

^aScalar coupling constants that were not reported previously due to extensive resonance overlap.^[1]

As the computational process involves the analysis of both the ¹H spin network of 1 and the total line shape of the spectrum, HiFSA is capable of identifying signal distortions that occur in higher-order coupled spin systems and can also reproduce the complex signal patterns encountered. As a result of the total line shape fitting, two small ⁵J (para) couplings within the Ph-l-threo-S residue [H5(o)-H8(m), H6(m)-H9(o)], one ⁵J coupling between the N-Me-4-OMe-l-W [H5(o) and H7(p)], as well as ⁴J (-meta) couplings within the Ph-l-threo-S [H5(o)-H9(o), H5(o)-H7(p), H7(p)-H9 (o), and H6(m)-H8(m)] and N-Me-4-OMe-l-W [H7(p)-H9(o)] residues were readily determined. These small couplings completed the J-coupling map of the aromatic spin systems of 1. The three signals, which had to be described previously as ‘broad singlet’ arising from the Ph-l-threo-S residue, could thereby be identified as belonging to an AA′BB′C spin system, with defined multiplicities and J_HH values. In addition, the ‘singlet’ nature of the signal associated with the N-Me-4-OMe-l-W residue can now be more accurately described as that of a doublet with a small coupling (J_HH = 0.48 Hz).

As the HiFSA-generated, numeric ¹H NMR profiles consist of the field-independent parameters, δ_H and J_HH, the corresponding high-resolution ¹H spectra (HiFSA fingerprints)^[6] can be simulated at any desired field strength and used for efficient NMR-based dereplication. To demonstrate the feasibility of this approach, Fig. 3 compares the ¹H NMR spectrum of 1 observed at 400 MHz with the simulated 400 MHz spectrum, calculated from the numeric HiFSA profile generated using the 900 MHz ¹H spectrum. The simulated HiFSA fingerprint spectrum is in excellent agreement with the lower-field experimental NMR data, displaying a calculated RMS of only 0.064%.

Figure 3.

Result of the ¹H iterative full spin analysis (HiFSA) of the 400 MHz ¹H NMR spectrum of ecumicin (1) (experimental [top] vs. simulated [bottom] spectrum).

Figure 4 shows sub-spectra of the individual amino acids generated from the ecumicin HiFSA profile. In this fashion, the severely overlapped experimental spectrum of 1 can be represented by the sum of the individual sub-spectra of its 13 amino acid residues. As expected, all l-V residues, including the two N-methylated valines, have very similar individual HiFSA fingerprints: The J_HH values are near identical, and the spectra are only subject to small Δδ_H effects arising from the different chemical environments within the molecule. This holds true for the two l-T residues (one l-T and one N-Me-l-T) as well. Although a large number of oligopeptides have been reported, the complexity and severe signal overlap in their ¹H NMR spectra have mostly resulted in limited information on the coupling pattern of their composing amino acids, whereas HiFSA deduces this information comprehensively as shown here.

Figure 4.

Sub-spectra of the individual amino acid residues of ecumicin (1), extracted from the ¹H iterative full spin analysis profile of the entire molecule.

Congruence was observed when comparing the J_H,H of l-V, l-T, and l-I of 1 to the available reported J_H,H values of these amino acids.^[6–11] These observations are consistent with the general assumption that the amino acid residues would essentially maintain their characteristic ¹H,¹H coupling patterns (J_HH values) when present in a different chemical environment, e.g. a different peptide. Based on this observation, we concluded that it should be possible to treat the ¹H spectrum of any new peptide as that of a mixture of their individual amino acids. This knowledge, taken together with the HiFSA profiles of individual amino acids, would then greatly assist in the structure elucidation and verification of any new peptide. Furthermore, as HiFSA is able to distinguish between even closely related molecules, as demonstrated recently for four (iso)silybin isomers, which exhibit near-identical ¹H NMR spectra,^[12] we expect that HiFSA can enable the identification of individual amino acids present in the ¹H spectrum of a peptide. Following this approach, the structures of several additional ecumicin-like analogs were elucidated and will be published separately in due course.

By using the simulation tools contained within PERCH and the characteristic parameters summarized in Tables 1 and 2, it is possible to estimate the field strength required to improve the resolution of the ¹H NMR spectrum of 1 up to a point at which it essentially becomes a first-order NMR spectrum: A ¹H observation frequency of approximately 4 GHz will be necessary to achieve such spectral resolution (Fig. S1). Although such an ultra-high-field instrument would certainly expedite the NMR analysis of small molecules with complex NMR spectra, the required magnetic field strength falls outside of what seems achievable with cryomagnetic technology in the foreseeable future. Until such advance systems can be developed, available tools for computer-aided NMR analysis, such as HiFSA, will continue to provide a means of understanding complex resonance patterns that will otherwise remain undefined.

In summary, possessing a molecular weight of 1599 amu and 138 protons, 1 is presently to our knowledge the largest molecule for which a ¹H full spin analysis has been achieved. A list of detailed and accurate δ_H and J_HH values is now available for 1, which permits the generation of the highly reproducible ¹H fingerprints of this anti-TB drug lead and its individual amino acid residues, at any field strength including lower-field instruments. Moreover, the HiFSA profiles of individual amino acids from a known peptide can be stored separately in nonproprietary text files (only 2–20 kB in size), and ultimately can assist in the simplification of the structure elucidation process of new peptides. Although a significant component of manual input is at present still required, this new methodology may represent the beginnings of a new paradigm and a valuable adjunct to fully automated NMR-based structure elucidation processes.

Experimental

NMR spectroscopy

A sample of ecumicin (1) was prepared by weighing 4.00 mg (±0.01 mg) into a microcentrifuge tube using a Mettler Toledo XS105 Dual Range analytical balance, followed by addition of 600 µL of methanol-d₄ (99.8%). The solution was transferred to a 5 mm NMR tube using a Pressure-Lok gas syringe (VICI Precision Sampling Inc., Baton Rouge, LA, USA). The final concentration of the ecumicin sample was 6.67 mg/ml. After acquiring a ¹H NMR spectrum at 400.13 MHz on a Bruker DPX-400 NMR spectrometer equipped with a 5-mm QNP (4-nucleus) probe, the solution was transferred back to a microcentrifuge tube and 200 µL of it was later transferred to a 3 mm NMR tube. A ¹H NMR spectrum was also measured at 899.94 MHz with a Bruker AVANCE-II-900 NMR spectrometer, equipped with 5-mm TCI triple resonance inverse detection cryoprobe and z-axis pulse field gradient. All ¹H NMR experiments were acquired at 298 K (25 °C) using the standard Bruker pulse sequence (“zg”), with qHNMR parameter optimization. The probes were frequency tuned and impedance matched prior to data collection. Chemical shifts (δ_H) are expressed in ppm relative to the residual protonated solvent signal (¹H 3.310 ppm), and coupling constants (ⁿJ_H,H) are given in Hertz (Hz).

Computer-assisted ¹H iterative full spin analysis

Full spin analysis was carried out with the PERCH NMR software tools according to previously published methodology.^[3] The resolution-enhanced ¹H NMR spectrum was imported into the PERCH shell as JCAMP-DX file and subjected to baseline correction, peak picking, and integration. The X-ray crystal structures of 1 (CCDC ID 940680)^[1,13] was used as a template to construct the molecular model of ecumicin using the PERCH molecular modeling system (MMS). Following geometry optimization and molecular dynamics simulations, the ¹H NMR parameters in methanol-d₄ were predicted using MMS. Following manual examination of the ¹H assignments, the calculated ¹H chemical shifts, signal line widths, and J-couplings were refined with PERCHit, using both the integral-transform (D) mode and the total-line-fitting (T) mode, until excellent agreement between the observed and simulated spectra was achieved to yield the ¹H fingerprints of ecumicin (RMS < 0.2%).