Enhanced measurement of residual chemical shift anisotropy for small molecule structure elucidation

A method is introduced to measure residual chemical shift anisotropies conveniently and accurately in the mesophase of poly-γ-(benzyl-L-glutamate). The alignment amplitude is substantially enhanced over common methods which greatly benefits measurements particularly on sp3 carbons. The approach offers significant improvements in data accuracy and utility for small molecule structure determination.


Introduction to the PBLG Liquid Crystal Medium
The phase transition and the texture of the cholesteric phase of PBLG have been thoroughly studied by microscopy and X-ray diffraction as described. [1][2][3] Briefly, at a low [PBLG], the solution is isotropic. When [PBLG] is higher than a critical concentration A, two phases co-exist in equilibrium including a heavier isotropic phase and a polymer-enriched mesophase characterized by a periodic multilayered supramolecular structure. Increasing [PBLG] beyond A converts more isotropic phase to mesophase with polymer concentrations in both phases staying constant, until [PBLG] reaches a second critical concentration B at which point the whole sample goes into mesophase; at an even higher [PBLG], the supramolecular packing density increases proportionally. The concentration gap between A and B is typically small for high-molecular weight PBLG and in this work we only utilize [PBLG] below A and above B for extraction of RCSA. If a biphasic solution is encountered, which can be revealed by an inhomogeneous 2 H spectrum of CDCl 3 showing one pair of signals separated by residual quadrupolar coupling (RQC) of several hundred Hz and a singlet or a pair of signals with barely resolved RQC, as described in ref 19 in the main text, a small amount of additional PBLG should be added until only the pair with a large RQC are observed.

Experimental Procedures Sample Preparation
Strychnine: Strychnine (30mg, MP Biomedicals, Inc.) was dissolved with 600 µL of CDCl 3 in a 5 mm NMR tube. Tetramethylsine (TMS) of 4% (v/v) was added for RCSA referencing. Carbon chemical shifts and 13 C-1 H J couplings were measured. Poly-γ-benzyl-Lglutamate (PBLG, 150,000-350,000 Da, Sigma-Aldrich catalog No.: P5136) was successively added to afford weight-to-volume ratios of 2.1%, 4.1%, 6.6%, 8.8%, 11.4%, 12.9%, 15.5%, 18.4%, 22.9%, and 34.5%, and after each addition, carbon chemical shifts were measured, as described in the main text. The 2 H spectrum of CDCl 3 was taken at each PBLG concentration to measure the residual quadrupolar coupling (RQC). The common method to dissolve PBLG is to briefly and repeatedly centrifuge the NMR tube with interleaved tube inversions, until a clear and homogenous solution was obtained. 4 Solid PBLG is lighter than CDCl 3 ; therefore by repeatedly inverting the NMR tube followed by centrifugation, the undissolved PBLG moves across the sample solution from one end to the other as it dissolves, creating a homogenous PBLG solution. See more details in the section of "Step-by-Step Procedure of NMR Sample Preparation".
Retrorsine: A sample of 10mg retrorsine (Sigma-Aldrich, Inc.) was prepared in the same fashion as for the strychnine sample, but with the successive addition of 0%, 8.1%, 12.6%, and 15.1% PBLG. TMS of 2.5% (v/v) was added for RCSA referencing. Carbon chemical shifts and CDCl 3 RQC were measured after each PBLG addition.
Caulamidine A: A sample of 1.6mg caulamidine A was dissolved in 170 µL CDCl 3 with 5 µL TMS in a 3mm NMR tube. The material was quantified by external calibration with another sample of 3.6 mg 1,3,5-trimethyoxybenzene (TMB) dissolved in 170 µL CDCl 3 , through comparison of the 1 H integration of caulamidine A or TMB over that of the residual CHCl 3 signal. RCSA was measured with I0 of no PBLG, I1 of 6.3% PBLG, and A1 of 13.1% PBLG. Carbon chemical shifts and CDCl 3 RQC were measured after each PBLG addition.

NMR Experiments
All NMR data were acquired on a Bruker 600 MHz spectrometer equipped with a helium cooled cryoprobe with the sample temperature maintained at 25 o C. For the measurement of carbon chemical shifts, the standard { 1 H}-13 C experiment was employed with a 90 o flip angle, a recycling delay of 2 s, and an acquisition time of 2 s, unless otherwise noted. Strychnine and retrorsine spectra were acquired with 32 and 64 transients, respectively, at all PBLG concentrations. As described in the main text, a higher decoupling field is needed to fully collapse large 13 C-1 H RDCs as the alignment gets stronger: for strychnine, proton decoupling was conducted using WALTZ16 with a decoupling field of 3125 Hz for the samples of 0%, 2.1%, 4.1%, 6.6%, and 8.8% PBLG in which the samples were isotropic; after alignment was induced at higher PBLG concentrations, proton decoupling was performed using WALTZ16 with a decoupling field of 4167 Hz for the sample of 11.9% PBLG; GARP with a decoupling field of 4167 Hz was used for the samples of 12.9%, 15.5%, and 18.4% PBLG, and GARP with a decoupling field of 8333 Hz was used for the samples of 22.9% and 34.5% PBLG. For retrorsine, proton decoupling was performed using WALTZ16 with a decoupling field of 3125 Hz for samples of 0% and 8.1% PBLG in which the samples were isotropic. GARP with a decoupling field of 4167 Hz was employed for the samples of 12.6% and 15.1% PBLG in which the samples were aligned. Carbon spectra of caulamidine A were collected with 800, 1024, and 3072 transients for 0%, 6.3%, and 13.1% PBLG, respectively. The acquistion time was 0.5 s for caulamidine A with 13.1% PBLG. Proton decoupling was performed with WALTZ16 with a decoupling field of 4167 Hz for 0% and 6.3% PBLG, and GARP with a decoupling field of 4167 Hz for 13.1% PBLG.

DFT Calculation
The density functional theory (DFT)-computed coordinates and chemical shielding tensors of strychnine and its twelve other diastereomers were adopted from a previous publication. 5 For retrorsine, the initial conformers were collected by combining outputs from three different conformer search programs including ET 6 , JG (a Merck in-house version of DG 7 ), and OpenEye's OMEGA 8 , and were further optimized by DFT in Gaussian 09 9 at the B3LYP/6-31G(d,p) level in the gas phase. The chemical shielding tensors were calculated in Gaussian 09 with the "nmr" keyword at the mPW1PW91/6-31G(d,p) level. 6. When the PBLG is fully dissolved, the solution will be clear.

Step-by-Step Procedure for NMR Sample Recovery
1. This sample recovery method is based on insolubility of PBLG in ethyl acetate (EtOAc). Conversely, the sample to be recovered must be soluble in EtOAc. If this is not the case, then chromatographic recovery (e.g., size exclusion chromatography) may be feasible. 2. The PBLG solution is quite viscous so it should be spun out of the tube using a centrifuge. First, cut a 0.5 in. piece of 3/16 in. I.D. Nalgene® tubing, which will fit snuggly around the 5 mm NMR tube. Drill a 0.5 cm hole in the top of a 15 mL conical bottom polypropylene centrifuge tube (Fisherbrand, P/N 05-539-12). Pre-rinse the centrifuge tube twice with EtOAc to remove any contaminants (e.g., plasticizer).
3. Insert the NMR tube into the centrifuge tube upside down and with the Nalgene® tubing "stopper" at the top such that the open end of the NMR tube just touches the beginning of the conical taper. Be sure to leave enough room below the open end of the NMR tube so that the sample solution spun out does not contact the tube opening.
4. Gently spin out the PBLG solution using a centrifuge. Rinse with 1-2 mL CDCl 3 followed by EtOAc to recover sample from the walls of the NMR tube, and continue to spin out each time using a centrifuge. 5. Concentrate the solution in the centrifuge tube under a stream of dry N 2 gas. Periodic agitation may be needed as the PBLG may form a film on top of the solution. 6. Add approximately 10 mL of EtOAc to precipitate the PBLG. The precipitation may be aided by vortexing. Further concentrate the solution down to 5 mL, which should fully remove residual CDCl 3 .
7. Rinse a 1.0 m PTFE filter (Whatman Puradisc ® 25 TF) and 5 mL plastic Luer-lock syringe with EtOAc to remove contaminants. 8. Centrifuge the precipitated solution in the centrifuge tube, and filter the supernatant into a 5 mL glass vial. Repeat by washing the filtrate 1-2 times more with EtOAc to ensure thorough sample recovery. 9. The EtOAc can then be simply removed under a stream of dry N 2 gas.

Summary of Sample Recovery Results
The sample recovery procedure was initially optimized on ibuprofen and then tested for seven additional small molecule compounds: fluconazole, acetanilide, lidocaine, trimethoprim, avobenzone, benzocaine, carbamazepine (all compounds purchased from Sigma-Aldrich as certified reference standards). Recovery was based on the insolubility of PBLG in ethyl acetate. Most small molecules that are soluble in chloroform are also soluble in ethyl acetate. However, in cases where the compound to be recovered is insolube in ethyl acetate, then chromatographic separation is recommended. The key to high recovery is ensuring that the viscous PBLG solution can be fully transferred between the NMR tube and centrifuge tube for precipitation with ethyl acetate. Several rinses with both chloroform and ethyl acetate were found to improve recovery from ~60% to >90%. Fig S1 shows the spectrum of ibuprofen before and after recovery. A small amount of extractable plasticizer (dioctyl phthalate) was also initially observed. This was traced to the plastic syringe barrel and could be minimized by pre-rinsing the syringe with ethyl acetate or by using glass Luer-lock syringes.
Simultaneous recovery of all eight tested compounds was found to be greater than 90% as measured by quantitative NMR. For the qNMR measurement, 1,3,5-trimethoxybenzene (TMB) was used as internal standard, and a relaxation delay of 60 s was used with a tip angle of 30°. The 6.1 ppm aromatic signal of TMB did not interfere with any sample component signals and was thus utilized for quantitation. The resolved signals of each compound used for quantitation were as follows: fluconazole, 8.06 ppm, s, 2H; acetanilide, 2.17 ppm, s, 3H; lidocaine, q, 4H; trimethoprim, 6.38 ppm, s, 2H; avobenzone, 7.97 ppm, d, 2H; ibuprofen, 7.21 ppm, d, 2H; benzocaine, d, 6.63 ppm, 2H; carbamazepine, 6.93 ppm, s, 2H. Figure S1. NMR spectra from the initial sample recovery test with ibuprofen. Note that the broad PBLG signal centered around 7.2 ppm completely disappeared after recovery. Although the recovery rate of this initial test was estimated to be only ~60%, the improved procedure achieved a 94.9% recovery rate (Table S1) by fully transferring the viscous PBLG sample from the NMR tube to the centrifugation tube.

The 1 H and 13 C spectra under Strong Alignment Conditions
With strong alignment in the PBLG mesophase, the homonuclear RDCs introduce an extensively coupled network and cause severe line-broadening of the 1 H spectrum. As shown in Fig S2a, in going from 0% to 8.1% PBLG, both of which were isotropic based on the absence of CDCl 3 RQC (Fig S2b), the increase in 1 H line-width was very modest. However, increasing the PBLG concentration to 12.6% caused a transition to the mesophase, as indicated by a CDCl 3 RQC of 272.6 Hz (Fig S2b), which was accompanied by substantial 1 H line-broadening due to the emergence of RDCs; such broadening was more pronounced with even stronger alignment at 15.1% PBLG (CDCl 3 RQC = 406.5 Hz). In contrast, as shown in Fig S2c, in the 13 C spectrum both the inhomogeneous broadening and homogenous broadening due to proton spin diffusion can be largely suppressed by proton decoupling, and therefore the line-width increase in PBLG mesophase is much less severe for 13 C than for 1 H. The sharp 13 C resonances faciliate accurate RCSA measurement even at high degrees of alignment. Also note that except for the two broad signals between 132 and 134ppm (Fig S2c expansion to the left), other background 13 C signals from the PBLG LC medium is extremely weak, yielding very clean spectra.

Rationale and Limitation of the  iso Correction Method
Our correction method is based on the premise that the  iso dependence on PBLG concentration is identical in both phases, such that the  iso in the mesophase can be corrected for by extrapolating its value from its trend in the isotropic phase.
This assumption is reasonable because of two points. First, the analyte can diffuse freely through the highly fluidic mesophase, 1-3 and second,  iso is determined to the first order by the total surface area of PBLG polymers from averaging all microstates of the analyte with orientational and positional degrees of freedom at the intermolecular interface. An effect not considered, however, is the interference between different surfaces on neighbouring PBLG molecules, which can become relevant with increasing PBLG concentration. This effect can, for instance, reduce the accessible area when two PBLG surfaces are very close or alter the microstates of an analyte trapped between two surfaces. It is noteworthy that the orientational redistribution of PBLG during the phase transition, while generating RCSA, is not expected to cause  iso by itself, although associated changes in surface interference may. However, as our studies in this work demonstrate, ignoring the interference does not cause significant errors to RCSA measurement.

Correction for  iso in RCSA measurement
As shown in Fig S3, the TMS-referenced isotropic carbon chemical shift ( 13 C  iso ) approximately follows a linear relationship as a function of PBLG concentration when the concentration is below the critical LC forming concentration. As described in the main text, this linear dependence forms the basis for the 13 C  iso correction used in the I0-I1-A1 method for RCSA measurement. Note that the 13 C  iso , as shown in Fig S3, is quite small in comparison to 13 C RCSA (See Table S1).

Effects of insufficient 1 H-decoupling in Strongly Aligned Samples
The potentially large and extensive 13 C-1 H RDCs observed for strongly aligned molecules may exceed the default 1 H decoupling capability commonly used for solution NMR experiments. As shown in Fig S4a (top trace), the default 1 H decoupling by WALTZ16 with a decoupling field of 3125 Hz led to the residual multiplet structure for C23 of strychnine aligned in 11.9% PBLG (Refer to Fig 1a in the main text for atomic numbering), whereas increasing the decoupling field to 4167 Hz collapsed the multiplet (Fig S4a, bottom trace); with an even stronger alignment at 12.9% PBLG, C23 showed up again as a residual multiplet with 4167 Hz WALTZ16 (Fig S4b, top trace). Switching from WALTZ16 to GARP yielded better off-resonance decoupling efficiency at the same decoupling power and restored the singlet (Fig S4b, bottom trace).