Application of Flow NMR to an Open-Access Pharmaceutical

Anal. Chem. 2003, 75, 1954-1957

Application of Flow NMR to an Open-Access Pharmaceutical Environment Gregory C. Leo,* Aaron Krikava, and Gary W. Caldwell

Johnson & Johnson Pharmaceutical Research and Development, LLC, Welsh and McKean Roads, P.O. Box 776, Spring House, Pennsylvania 19477-0776

We present the application of flow NMR in an automated, open-access environment. The adjustment of parameters affecting the selection of the correct sample size, the elimination of carry-over, and the optimization of sample recovery are addressed. Advantages of this method include ease of use, elimination of NMR tubes, and elimination of handling errors that can result in the contamination of the probe. Sample throughput is similar to instruments using a conventional autosampler (for Bruker instruments, a BACS) although the time used for shimming the sample can be eliminated when the sample solvent does not change. The key feature of our methodology is that only one push solvent is used. This has major advantages over methods that switch solvents because there is no extensive flushing required between solvents and the deuterated push solvent, deuterium oxide, is economical. The disadvantage is the need to maintain the push and transfer solvents and usually the loss of the exchangeable protons that the sample may have. The protocol we present, using a single push solvent, contributes to the application of flow NMR in hands-on medicinal chemistry environments. With the advent of rapid, automated techniques for screening and synthesizing compounds in the pharmaceutical industry, the workhorse analytical tools (NMR and MS) have experienced many new developments and enhancements. Desktop mass spectrometers with automated sample loaders are commonplace. Commercial mass spectrometers with up to eight inlet ports to increase sample throughput are now available (MUX-technology by Micromass). Hyphenated methods such as HPLC NMR1-3 and HPLC NMR MS1,4 are found in most development and metabolism groups. NMR spectrometers have for years had autosamplers that handled glass NMR tubes, but now with the development of a new generation of flow probes, liquid sample injection has found additional applications. These applications include screening for ligands bound to receptors,5,6 obtaining spectra from 96 well * Corresponding author. E-mail: [email protected]. Fax: 215-628-7064. (1) Albert, K. On-line LC NMR and Related Techniques; John Wiley and Sons: New York, 2002. (2) Lindon, J. C.; Nicholson, J. K.; Wilson, I. D. Adv. Chromatogr. 1996, 35, 315-82. (3) Lindon, J. C., Nicholson, J. K.; Sielmann, U. G.; Wilson, I. D. Drug Metab. Rev. 1997, 29, 705-746. (4) Lindon, J. C.; Nicholson, J. K.; Holmes, E.; Everett, J. R. Concepts Magn. Reson. 2000, 12, 289-320.

1954 Analytical Chemistry, Vol. 75, No. 8, April 15, 2003

plates,7 and profiling of bodily fluids for metabonomics studies.8 These applications all have the goal of increasing throughput, and they all share a common feature. This common feature is that the samples being pumped into the probe are all homogeneous; for example, they are all urine samples or medicinal chemistry compounds dissolved in the same solvent or aqueous samples of a receptor and ligands. An application not cited is the use of liquid sample injection for routine walkup use by chemists with different samples in a variety of solvents. This application is the topic of this report. Our work illustrates how flow injection offers cost savings because it eliminates NMR tubes and reduces samplehandling errors around the magnet that can possibly lead to contamination, downtime, or both. EXPERIMENTAL SECTION Experiments were performed using a Bruker DMX-400 with a 9.39-T unshielded magnet in conjunction with a 5-mm single-cell flow probe (0.0240 mL active volume, ∼0.500 mL total NMR cell volume) and a Gilson 215 liquids handler (chip version 2.30) with a septum-piercing needle. Chip version 2.3 resulted in smoother action of the diluter syringe and eliminated dispersion seen at the ends of the sample volume. The spectrometer was interfaced to a Silicon Graphics Indy computer (IRIX 6.5) using Bruker’s XWINNMR 3.0 and ICONNMR 3.0. The BEST_215.XCP software was used to interface the Gilson 215 to the Bruker NMR spectrometer. We used Bruker’s standard configuration of the Gilson 215 that consisted of two six-way valves, one for sample loading and one for sample injection into the probe. We used deuterated water for the push solvent. The push solvent is the solvent used to push the sample after it has been drawn up into the solvent loop to the NMR flow probe. A Valvemate was obtained for this system so that we could match the push solvent with the sample solvent, but we found it impractical because of the excess time needed for sample switching and the much higher costs for other deuterated solvents. The transfer solvent is the solvent used to form solvent plugs or gaps around the sample to be analyzed. The transfer solvent is the same solvent used for solubilizing the analyte, and both the transfer and sample solvents are deuterated. (5) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science 1996, 274, 1531-1534. (6) Stockman, B. J.; Farley, K. A.; Angwin, D. T. Methods Enzymol. 2001, 338, 230-246. (7) Kiefer, P. A.; Smallcombe, S. H.; Williams, E. H.; Salomon, K. E.; Mendez, G.; Belletire, J. L.; Moore, C. D. J. Comb. Chem. 2000, 2, 151-171. (8) Nicholson, J. K.; Lindon, J. C.; Holmes, E. Xenobiotica 1999, 29, 11811189. 10.1021/ac026389l CCC: $25.00

© 2003 American Chemical Society Published on Web 03/11/2003

The samples are placed in 1.8-mL sample vials with septum caps and placed in a Gilson 209 rack. The solvent and gas gaps used for introducing the sample into the probe and flushing the system are described in the next section. A general diagram of the Bruker flow-NMR system has been published.9 RESULTS AND DISCUSSION Having the goal to present the flow injection NMR as a walkup tool for synthetic medicinal chemists, it was desirable to have the spectral results look as similar to those obtained from an instrument using a sample robot in the automation mode. This necessitated the use of deuterated solvents as the push and transfer solvents to avoid diminution of the spectral window from solvent suppression methods. Deuterium oxide was chosen as the push solvent because it is inexpensive (∼$300/L) and is easily handled, reducing solvent safety concerns. The negative aspect when deuterium oxide is used is that exchangeable protons from the sample will usually be lost. The loss of exchangeable protons is generally not a problem for routine sample evaluation. The transfer solvents are the typical deuterated solvents commonly used for sample preparation. The sample to be analyzed is prepared in any deuterated solvent of choice with the exception of DMSO-d6, which we avoid because DMSO-d6 can freeze in the laboratory and this would cause sample-transfer problems. If samples must be solubilized in DMSO-d6, we found that the sample can be dissolved in DMSO-d6 and then diluted with D2O such that DMSO is ∼5 vol %. In this manner, the sample can be run as a D2O sample. Recently, heated transfer lines have become available from the instrument vendors and thus avoid the issue of DMSO solidifying. The standard method we have chosen for introducing samples into the NMR flow probe is as follows:

push solvent - (gas - solvent)3 gas - sample - gas f to probe D2O (0.025 - 0.010)3 0.025 0.600 0.025

The gas used is house nitrogen, and the unit of measure for the numbers listed is milliliters. The “solvent” in parentheses refers to the transfer solvent, and “sample” refers to the sample dissolved in the appropriate solvent that is matched by the transfer solvent. Watching the Gilson 215, what one will see is that three aliquots of transfer solvent are drawn from the transfer bottle always bracketed by gas spaces and then the sample is drawn up. The entire train of solvent and gas plugs is then pushed to the probe with the sample at the front. Using deuterium oxide as the push solvent in conjunction with organic sample solvents might lead the reader to expect that there may be susceptibility problems during autoshimming. This is not a problem. The samples can be readily shimmed either manually or in the automation mode. We have reference shim files for the different solvents, and in automation, a quick shim (one iteration) of Z1 and Z2 is sufficient when the sample solvent is changed. The reference shim files are generated manually and take into account any solvent susceptibility issues. The main susceptibility (9) Spraul, M.; Hofmann, M.; Ackerman, M.; Nicholls, A. W.; Damment, S. J. P.; Haselden, J. N.; Shockcor, J. P.; Nicholson, J. K.; Lindon, J. C. Analyst (Cambridge, U.K.) 1997, 122, 339-341.

Figure 1. Spectra showing the effects of the injected sample volume upon the line shape. Only the aromatic region of acetylsalicylic acid (aspirin) dissolved in CD3OD is shown. The top spectrum was from a sample volume equal to 0.45 mL, and a sample volume of 0.60 mL yielded the bottom spectrum.

problem to overcome is between air and solvent. This was only a problem initially when the system was set up and a sample volume insufficient to fill the entire flow cell was used (Figure 1). The top spectrum shows the effect of susceptibilility line broadening because the sample chamber is not completely filled. Under such conditions as these, it is not possible to shim the sample. The bottom spectrum using the same sample as in the top spectrum but filling the sample chamber completely resulted in sharp resonances. The resonances shown are the aromatic peaks of aspirin dissolved in methanol-d4, and the resolution obtained represents what is normally achievable using flow injection NMR in conjunction with automated data collection and processing. We chose a sample volume that allows for variation in sample placement within the probe coil so that air bubbles will not form in the sample chamber. To understand the problem of not adequately filling the sample chamber, we removed the outer casing of the probe so that we could see when the sample chamber was filled using a syringe. For our particular flow probe, the sample chamber was ∼0.5 mL and the volumes of the lines to and from the probe are ∼0.1 mL. We found 0.6 mL was enough volume to reliably fill the sample chamber and leave enough of a margin for variation among the solvents as to where the solvent front actually stops when it is pumped to the probe. Representative spectra acquired using automation for samples dissolved in common solvents are presented in Figure 2 and are displayed so as not to clip any of the peaks vertically. The spectra are acquired without any solvent suppression using a standard pulse and acquisition-type sequence. It can be seen that the push solvent does not interfere or reduce the quality of the data. The amount of sample preferred is 1-10 mg, which is enough to give good proton spectra using 16 transients on our 400-MHz spectrometer. It is desirable not to use larger samples to minimize the time needed to flush the probe to eliminate carry-over from sample to sample. The top spectrum in Figure 2 is of caffeine (∼7 mg/mL) dissolved in deuteriochloroform. It is a very simple spectrum having one aromatic proton resonance and three N-methyl signals. The residual proton signal from the chloroform is the second peak from the left. The reference signal for TMS is seen at the far right of the spectrum, and the residual water in the chloroform (plus any water signal from the push solvent) is the very small peak midway between TMS and the N-methyl Analytical Chemistry, Vol. 75, No. 8, April 15, 2003

1955

Figure 2. Spectra obtained using flow injection NMR from samples dissolved in typical solvents. All the spectra are shown on-scale. The top spectrum is from caffeine (∼7 mg/mL CDCl3); the middle spectrum is from aspirin (2 mg/mL CD3CN), and the bottom spectrum is from aspirin (2 mg/mL CD3OD). In all cases, D2O is used as the push solvent. An “r” is used to indicate the residual water peak, and an “s” denotes the solvent peak. Tetramethylsilane (TMS) is seen at the right in the top and bottom spectra.

resonances. The middle spectrum of Figure 2 shows aspirin (2 mg/mL) dissolved in deuterioacetonitrile, and again it is seen that the push solvent presents no obfuscation of the solute resonances. The upfield resonance is the pentuplet of deuterioacetonitrile, and the residual water peak (∼2.3 ppm) is the foot observed on the low-field side of the acetyl peak of aspirin. There are two doublets and two triplets in the aromatic region. The carboxyl proton of aspirin is not seen in the downfield region of the spectrum (not shown) because of solvent exchange with the deuterium oxide (the push solvent). The bottom spectrum is aspirin dissolved in CD3OD, and again the data are readily interpretable and there is no loss of information because of large solvent peaks or solvent suppression techniques. It should be noted that the residual water resonance might move or have a broad peak shape especially as the D2O push solvent picks up moisture from the air. It is desirable to keep the D2O relatively fresh to prevent contamination by atmospheric water. Figure 3 shows some spectra obtained while attempting to minimize sample carry-over. The top spectrum is from caffeine dissolved in methanol-d4 (7 mg/mL). Caffeine’s sharp methyl singlets provide a good test of protocols for eliminating carryover. The second spectrum from the top is for a less concentrated solution of acetylsalicylic acid in methanol-d3 (2 mg/mL) injected after the caffeine sample. This spectrum contains carry-over peaks (marked with asterisks) from the previous caffeine sample. The small peaks observed in the aromatic region are impurities present in the sample. The wash cycle used is the one described below minus the second flush of the transfer lines to and from the probe. In the third spectrum from the top, these carry-over peaks were 1956 Analytical Chemistry, Vol. 75, No. 8, April 15, 2003

Figure 3. Spectra showing the elimination of carry-over signals. Spectra in order from the top: caffeine (∼7 mg/mL CD3OD); aspirin (2 mg/mL CD3OD) exhibiting carry-over peaks (one rinse of the transfer lines); aspirin (2 mg/mL CD3OD) without carry-over peaks (two rinses of the transfer lines); and aspirin (2 mg/mL CD3CN) without carry-over (again with two rinses of the transfer lines). The asterisks mark the carry-over peaks. Each of the aspirin spectra was acquired after an injection of the caffeine sample.

eliminated entirely by repeating the wash step of the lines leading to the probe. The bottom spectrum is the result obtained when solvents are changed, and again it can be seen that carry-over is eliminated. The bottom two spectra were each preceded by injections of the caffeine sample in methanol. Our cleaning protocol between samples involves cleaning the needle with 0.250 mL of transfer solvent. The transfer solvent is preceded by a 0.250-mL nitrogen gas gap. The outside of the needle is rinsed with 0.300 mL of transfer solvent. Then the lines to and from the probe are flushed with 0.300 mL of transfer solvent (the flush is preceded by 0.300 mL of nitrogen gas), and this step is repeated twice. We found the second rinse to be a key step that enabled rapid solvent changes with a minimal increase in the use of deuterated solvent. The final cleaning step is a 1.500-mL wash of the probe using the push solvent followed by a nitrogen gas purge for 45 s. Samples are returned to separate vials in a separate rack mirroring the position from the original rack. Sample recovery is ∼85% and is accomplished by syringe action pulling the sample along with N2 pressure pushing the sample. With our version of software, it was not possible to recover more of the sample even though sample could still be observed in the transfer line. Between samples, the previously described cleaning protocol is executed. In comparing the automated flow injection method to the automated glass NMR tube method, the use of deuterated solvent is greater for the flow injection method since an ampule of deuterated solvent (0.8 mL) is used for the sample, 0.03 mL is used for sample transfer (i.e., the solvent gaps), 0.85 mL is used for needle and system washing, and ∼4 mL of D2O is used as the push solvent. Using methanol for cost comparison, the cost per sample using flow NMR is $4.29 ($300/1000 mL × 4 mL of D2O + $1.90/ampule methanol-d3 + $15.20 bulk price/10 g × 0.89

g/mL × 0.88 mL methanol-d3) and for traditional automation using NMR tubes the cost is $5 ($1.90/ampule of methanol-d3 + $3.10/ NMR tube). The cost is slightly lower for the flow method, and in an environment where 7500 samples/month are run, there could be savings of approximately $5000. The cost difference would more significantly favor the flow NMR method if a smaller volume flow probe or a capillary flow probe was used along with a shielded magnet, which would reduce the length of the transfer line we use by a factor of ∼2. These types of changes would reduce the expenses incurred from deuterated solvents. Time comparisons between the two methods, in our hands, are about the same, 5-6 min/sample when the same number of scans per sample is used. When samples are run in the same solvent, the shimming step is skipped, reducing the time to run the sample by 1-2 min (for flow injection NMR). The flow NMR method has the advantage of being a closed system; that is, sample tubes are not going in and out of the probe, completely avoiding sample breakage and the introduction of dirt from the environment onto the spinner bearing and the glass probe inserts. The disadvantage of the flow system is the necessity to maintain the solvent reservoirs and loss of exchangeable protons in the samples. Additionally, our system is limited in that we cannot reliably use samples dissolved in pure DMSO. The training of the users’ group for the instrument should include a warning about the unannounced, rapid movement of the arm of the liquids handler (Gilson 215). As the vendors continue to develop and improve the technology, it may soon be the common mode of operation in openaccess industrial environments for its ease of use and cost savings. Our protocol using a single push solvent will contribute to the application of flow NMR in hands-on environments.

CONCLUSION Our goal was to develop flow injection NMR as a tool for chemists in a hands-on environment. To this end, the sample volume must be adjusted for the flow probe so that the sample cell is completely filled, thus avoiding air/solvent interfaces proximal to the probe receiver coil. The removal of this type of susceptibility problem is required before optimization of sampletransfer protocols can be successfully negotiated. The choice of a single push solvent, deuterium oxide, is both practical and costeffective and makes feasible the acquisition of data that are appropriate for medicinal chemists even though deuterium oxide will usually exchange with the exchangeable protons in the sample. Carry-over can be readily controlled in a manner that is not time-consuming but requires some limits upon the quantity of sample injected into the probe. In a medicinal chemistry environment, 10 mg is a reasonable limit and is an amount that can be readily flushed out of the system prior to the introduction of the next sample. The process of optimizing the flow injection probe was timeconsuming, at least for the generation of software that we have, because of the necessity of the liquids handler to initialize all of its parameters after a single change was made. In this respect, it suffers in comparison with the traditional NMR tube autosamplers (specifically, the Bruker BACS with which we have familiarity) since once installed by the vendor, they are essentially ready to be used. Received for review December 3, 2002. Accepted February 10, 2003. AC026389L

Analytical Chemistry, Vol. 75, No. 8, April 15, 2003

1957