Anal. Chem. 2000, 72, 3922-3926
Reversed-Phase High-Performance Liquid Chromatography Combined with On-Line UV Diode Array, FT Infrared, and 1H Nuclear Magnetic Resonance Spectroscopy and Time-of-Flight Mass Spectrometry: Application to a Mixture of Nonsteroidal Antiinflammatory Drugs Dave Louden and Alan Handley
LGC, The Heath, Runcorn, Cheshire,WA7 4QD, U.K. Steve Taylor, Eva Lenz, Steve Miller, and Ian D. Wilson*
Zeneca Pharmaceuticals, Mereside, Alderley Park, Macclesfield, Cheshire SK10 4TG, U.K. Ashley Sage
Micromass Ltd., Floats Road, Altrincham, Wythenshawe, Manchester M23 9LZ, U.K.
A prototype multiply hyphenated system has been applied to the analysis of a mixture of nonsteroidal antiinflammatory drugs separated by reversed-phase HPLC. Characterization of the model NSAIDs was achieved via a combination of diode array UV, 1H NMR, FT-IR spectroscopy, and time-of-flight mass spectrometry. This combination of spectrometers allowed the collection of UV, 1H NMR, IR, and mass spectra together with atomic composition data enabling almost complete structural characterization to be performed. We have been interested for some time in the possibilities of characterizing compounds present in complex mixtures via the multiple hyphenation of the required spectroscopies in order to provide comprehensive spectroscopic information in a single analysis. In particular, together with a number of other groups, we have actively investigated the utility of chromatography doubly hyphenated with nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) for this purpose (e.g., see refs 1-9, reviewed in ref 10). Such systems provide a very powerful * Corresponding author: (tel) 00 44 1625 513424; (fax) 00 44 1625 583074; (e-mail)
[email protected] . (1) Pullen, F. S.; Swanson, A. G.; Newman, M. J.; Richards, D. S. Rapid Commun. Mass Spectrom. 1995, 9, 1003-1006. (2) Shockor, J. P.; Unger, S. E.; Wilson, I. D.; Foxall, P. J.; Nicholson, J. K.; Lindon, J. C. Anal. Chem. 1996, 68, 4431-4435. (3) Holt, R. M. Newman, M. J.; Pullen, F. S.; Richards, D. S.; Swanson, A. G. J. Mass Spectrom. 1997, 32, 64-70. (4) Clayton, E.; Taylor, S.; Wright, B.; Wilson, I. D. Chromatographia 1998, 47, 264-270. (5) Scarfe, G. B.; Wright, B.; Clayton, E.; Taylor, S.; Wilson, I. D.; Lindon, J. C.; Nicholson, J. K. Xenobiotica 1998, 28, 373-388. (6) Dear, G. J.; Ayrton, J.; Plumb, R.; Sweatman, B. C.; Ismail, I. M.; Fraser, I. J.; Mutch, P. J. Rapid Commun. Mass Spectrom. 1998, 12, 2023-2030. (7) Wilson, I. D.; Morgan, E. D.; Lafont, R.; Shockor, J. P.; Lindon, J. C.; Nicholson, J. K.; Wright, B. Chromatographia 1999, 49, 374-378.
3922 Analytical Chemistry, Vol. 72, No. 16, August 15, 2000
means for structure determination, and indeed, the combination of retention time with NMR and MS spectral data is often enough to enable the unequivocal identification of an analyte. However, other spectroscopic techniques can also be linked to chromatographic separations that can be used to provide additional structural information. Thus, ultraviolet (UV) spectra can also provide supportive data for confirmation of identity and infrared (IR) spectroscopy can aid in the identification of structural features such as carbonyl or nitrile groups. In previous work, we have examined the practicality of coupling of size exclusion chromatography (SEC) with on-flow NMR and on-line collection of peaks via a dedicated interface for subsequent off-line FT-IR spectroscopy.11 A further refinement of this system was provided by the subsequent addition of on-line MS.12 However, while this system did enable NMR, MS, and IR spectra to be obtained for components of simple mixtures separated in a single chromatographic experiment, the need to obtain the FT-IR spectra off-line did make the methodology somewhat cumbersome. The experience gained with this system did however show the potential for multiple spectroscopic characterization, and more recently, we constructed a flow injection analysis system (FIA) for the characterization of compounds from, for example, chemical libraries.13 This prototype system consisted of a UV diode array (8) Scarfe, G. B.; Wright, B.; Clayton, E.; Wilson, I. D.; Lindon, J. C.; Nicholson, J. K. Xenobiotica 1999, 29, 77-91. (9) Hansen, S. H.; Jensen, A. G.; Cornett, C.; Bjornsdotitir, I.; Taylor, D.; Wright, B.; Wilson, I. D. Anal. Chem. 1999, 71, 5235-5241. (10) Wilson, I. D. J. Chromatogr., A, in press. (11) Ludlow, M.; Louden, D.; Handley, A.; Taylor, S.; Wright, B.; Wilson, I. D. Anal. Commun. 1999, 36, 85-87. (12) Ludlow, M.; Louden, D.; Handley, A.; Taylor, S.; Wright, B.; Wilson, I. D. J. Chromatogr., A 1999, 857, 89-96. (13) Louden, D.; Handley, A.; Taylor, S.; Lenz, E.; Miller, S.; Wilson, I.; Sage, A. Analyst 2000, 125, 927-932. 10.1021/ac000204y CCC: $19.00
© 2000 American Chemical Society Published on Web 07/19/2000
Figure 1. Experimental layout of the various spectrometers used in this multiply hyphenated HPLC-UV(DAD)-FT-IR-NMR-TOF-MS system.
detector (DAD), a FT-IR spectrometer, an NMR spectrometer, and a time-of-flight (TOF) MS (providing the ability to determine molecular formulas via accurate mass determination). Having successfully applied this multiply hyphenated system to FIA, we then sought to add a further dimension to the work by including a chromatographic separation to allow the analysis of mixtures. Here the application of an HPLC-UV(DAD)-FT-IR-NMR-TOFMS system to the separation and characterization of a simple mixture of nonsteroidal antiinflammatory drugs is described. EXPERIMENTAL SECTION Reagents. The compounds employed in this investigation were ibuprofen, flurbiprofen, naproxen, and indomethacin (Sigma, Poole, U.K.) (structures inset to Figures 3-6). Samples were dissolved in deuterium oxide (D2O) to give a sample containing 10 mg mL-1 each of the analytes. The HPLC system (Figure 1 consisted of a Bruker LC22 pump (Bruker, Coventry, U.K.), which delivered eluent at 1 mL min-1. Typically, 200 µL of sample was introduced into the flowing stream via a model 7125 Rheodyne injector fitted with a 200-µL sample loop. Chromatography was performed using a Hichrom 5-µm C18 bonded column (HiRPB 150 × 4.6 mm i.d.). The HPLC mobile phase consisted of acetonitrile (Pestanal Grade, Riedl de Haen) and D2O (99.8% isotopic purity, Fluorochem) at a 1:1 v/v ratio, containing 1% deuterated formic acid, 99.8% isotopic purity (Cambridge Isotope Laboratories) to give a pH of ∼2. On emerging from the column, the eluent flowed into a Varian 9065 UV diode array detector (Varian UK Ltd., Surrey, U.K.) via 30 cm of 0.005-in.-i.d. polyethyl ether ketone (PEEK) tubing. UV
spectra were collected over the wavelength range 190-360 nm, using the Star Chromatography workstation, version 4.0 (Varian UK Ltd.), and analyzed for spectral information using Polyview version 2.0 (Varian UK Ltd.). Following DAD the flow went via 110 cm of 0.005-in.-i.d. PEEK tubing to a Bio-Rad (Cambridge, MA) FT-IR model 375C spectrometer fitted with a Spectra Tech (Stamford, CT) Macro Circle Cell attenuated total reflectance (ATR) stainless steel flow cell of 400-µL volume fitted with a zinc selenide ATR crystal. Spectra were acquired with the kinetics software collecting 20 scans/ spectrum (5-s acquisition time) with a sensitive mercury cadmium telluride (MCT) liquid nitrogen-cooled detector. The spectra were acquired at 8-cm-1 spectral resolution. The ratio of the sample spectrum to that of a background spectrum of the flowing solvent through the cell was determined prior to injection of the sample solution, thus automatically subtracting out the solvent spectrum from the sample spectra. Following FT-IR the flow entered, via 150 cm of 0.005-in.-i.d. PEEK tubing, a Bischoff Lambda 1000 UV detector (Bruker, Coventry, U.K.) set at 254 nm after which the solvent stream was split 95:5 with 5% of the flow being directed to the mass spectrometer via 250 cm of 0.007-in.-i.d. PEEK tubing and the remainder to the NMR via 280 cm of 0.01-in.-i.d. PEEK tubing. Further details concerning the layout of this system are shown in Figure 1. Mass spectra were acquired on a Micromass LCT TOF mass spectrometer (Micromass, Altrincham, U.K.) using electrospray ionization (ESI) with a Z Spray source. The nebulizer gas flow Analytical Chemistry, Vol. 72, No. 16, August 15, 2000
3923
was set to 85 L h-1 and the desolvation gas to 973 L h-1. Spectra were acquired in either positive or negative ion mode with a capillary voltage of 3.2 kV and a cone voltage of 25 V. The source temperature was set to 120 °C and the desolvation temperature to 350 °C. The pusher cycle time was 50 µs with 0.9-s acquisitions and an interacquisition delay of 0.1 s. The mass range was 100900 Da. Diclofenac, molecular mass 295.0152 Da, was used to provide a lock mass for the spectrometer in negative ion mode and was introduced via a t-piece at 0.5 mL min-1 at a concentration of ∼5 ng mL-1. The exact concentration of diclofenac was adjusted to give approximately 300-500 counts signal intensity to allow good accurate mass confirmation. All of the instrumentation described above was located outside the 5 G line of the stray magnetic field generated by the 500MHz NMR spectrometer. NMR spectra were acquired using a Bruker DRX-500 NMR spectrometer in the stop-flow mode at 500.13 MHz using a flowthrough probe of 4-mm i.d. with a cell volume of 120 µL. Double solvent suppression was achieved via the NOESYPRESAT pulse sequence (Bruker Spectrospin) with irradiation of the acetonitrile (ACN) and residual water signals during the relaxation delay of 2 s and the mixing time of 100 ms. FIDs were collected into 16K data points over a spectral width of 8278 Hz, resulting in an acquisition time of 0.99 s. Between 32 and 124 scans were acquired for each spectrum. 1H NMR spectra were internally referenced to ACN at 1.93 ppm. RESULTS AND DISCUSSION The instrument layout shown in Figure 1 used in this prototype multiply hyphenated HPLC system was based on the design arrived at during our investigation of the FIA system. As noted for the FIA system, it should prove be possible to use different configurations in future systems and thereby obtain improved performance. However, the configuration used here with chromatography followed first by UV-DAD, FT-IR, and then the flow split 95:5 between the NMR and TOF-MS respectively proved satisfactory for the experiments described here. Chromatography. The chromatographic system used a C18 bonded stationary phase in combination with a simple reversedphase isocratic solvent system formed from a 1:1 mixture of D2O and acetonitrile containing 1% formic acid to ensure an acidic pH (∼2). This was necessary because all four of the NSAIDs used to prepare the mixture were acidic and had to be maintained in the un-ionized form to ensure chromatographic retention. The separation achieved with this system, monitored at 254 nm using the variable-wavelength UV detector placed after the FT-IR instrument (see Figure 1), is shown in Figure 2. This UV detector was also used to trigger the acquisition of stopped-flow NMR spectra. Despite the relatively high sample loading used (∼2 mg/ compound), good chromatographic peak shapes were obtained with baseline separation for all four compounds with a total analysis time of ∼17 min. An attempt to use of higher concentrations of acetonitrile (60%) to shorten the analysis time resulted in the coelution of indomethacin and ibuprofen. The small peak observed for ibuprofen compared to the other three test compounds is a result of the low UV absorption of the compound at 254 nm rather than an indication of a lower amount of this analyte injected. 3924 Analytical Chemistry, Vol. 72, No. 16, August 15, 2000
Figure 2. HPLC-UV (254 nm) chromatogram of the separation of the test mixture of nonsteroidal antiinflammatory drugs. Key: 1, naproxen: 2, flurbiprofen: 3, indomethacin; 4, ibuprofen.
Figure 3. TOF-MS, UV, IR, and 1H NMR spectra of naproxen (structure in inset to UV spectrum).
UV Spectra. Although the quantities of material on column were sufficient to overload the UV-DAD it was nevertheless possible to obtain characteristic UV spectra from the leading and tailing edges of the individual peaks. These spectra, shown as insets to Figures 3-6, compared well with previously obtained spectra for standards. Although the limitations of UV spectroscopy as a means of identification/confirmation of identity are well recognized, these spectra nevertheless provide further evidence,
Figure 5. TOF-MS, UV, IR, and 1H NMR spectra of indomethacin (structure in inset to UV spectrum). Figure 4. TOF-MS, UV, IR, and 1H NMR spectra of flurbiprofen (structure in inset to UV spectrum).
to add to chromatographic retention data, that the peaks do indeed correspond to the compounds present in the mixture. In future investigations, it would, however, be beneficial to place the UVDAD on the branch of the flow taking the minor portion of the eluent (5%) to the mass spectrometer and use a makeup flow to dilute the sample in order to bring the whole of the peak onscale. This would then enable UV spectra to be obtained for the whole peak. IR Spectra. On-flow spectra were obtained for all of the analytes in the mixture as shown in Figures 3-6. As shown in these figures, the spectra of all four NSAIDs were dominated by the carbonyl absorption at ∼1700 cm-1, typical of a carboxylic acid function. The analytes are distinguishable from each other on the basis of the considerable differences between compounds in the “fingerprint” region of the infrared spectrum, while the signal-to-noise ratio was quite adequate for identification for all of the compounds except ibuprofen, where it was borderline. As well as visual inspection, the spectra were also searched against a spectral library containing over 6000 spectra (including those of the test compounds). In each case, the top match corresponded to the correct analyte. In the case of naproxen, flurbiprofen, and indomethacin the spectral matches were 83.7, 84.5, and 92.4%, with the second match for these compounds being 53.9, 45.9, and 63.2%, respectively. For ibuprofen, the top match
was also correct but in this case at only 66.7%; with the second match at 66.3%, identification was less clear-cut. This was due to the relatively poor signal-to-noise ratio obtained for this compound compared to the others. Improvements in the signal-tonoise ratios could be achieved by coaddition of all of the spectra obtained over the eluting chromatographic peaks, or alternatively, a lower flow rate could be employed to enable more spectra to be obtained. NMR Spectra. In our previous studies using this multiply hyphenated system for FIA.13 NMR spectra were obtained onflow. However, in this application, to ensure that good quality spectra were obtained and to minimize the difficulties of solvent suppression, the stop-flow mode was used. Thus, the 1H NMR spectra shown in Figures 3-6, acquired for each of the model compounds, were obtained using between 32 and 124 scans/ spectrum. The spectra contain, in addition to the resonances for the analytes themselves, signals for residual water (4.1 ppm), acetonitrile (1.93 ppm), methanol (3.2 ppm), and the formic acid (8.1 ppm) used for suppression of the ionization of the analytes. However, none of these additional resonances prevented identification of the signals of the NSAIDs present in this mixture. Thus, in the case of the first-eluting component naproxen, all of the aromatic resonances are clearly visible between 7.0 and 7.8 ppm, together with the singlet for the methoxy group (3.82 ppm) and doublet for the aliphatic methyl (∼1.4 ppm) (see Figure 3). The quartet for the methine proton coupled to the alihatic methyl is also visible at ∼3.81 ppm albeit partially obscured by Analytical Chemistry, Vol. 72, No. 16, August 15, 2000
3925
formic acid, respectively. Essentially similar results were seen with indomethacin (peak 3, Figure 2) and ibuprofen (peak 4, Figure 2) with molecular ions, [M - D]-, at m/z 356.0668 (calcd 356.0690, C19H15NO4Cl) and 205.1252 (calcd 205.1229, C13H17O2), respectively. Also present in each spectrum, to varying degrees, was the lock mass (m/z 295). Due to the high sample loadings of the analytes employed to ensure the acquisition of NMR and IR spectra, there was an excess of compound reaching the mass spectrometer, which reqiured the use of a large split, and a makeup flow in order to reduce this. Thus, the concentration of the compounds was such that only a few acquisitions could be summed together and this coupled with the suppression of the lock mass signal meant we were not able to get within the expected specification for the instrument (better than 2 mDa below mass 400). In a fully optimized system, we would have used a much larger accurate fixed split, rather than the t-piece employed here, and we would have optimized the makeup flow to deliver the best possible composition of sample and lock mass to the spectrometer. Even so, we were able to get accurate mass data to less than 12 ppm for all the compounds in our mixture. Further, when the software to assign molecular compositions to the masses based on these data was used, the correct compound was the first hit in each case.
Figure 6. TOF-MS, UV, IR, and 1H NMR spectra of ibuprofen (structure in inset to UV spectrum).
the resonance for the methoxy group. Similarly, all of the expected resonances for the two remaining “profen”-type NSAIDs (flurbiprofen and ibuprofen) can be observed in the 1H NMR spectra obtained in this experiment (see Figures 4 and 6). Indomethacin (peak 3 in Figure 2) is structurally dissimilar to the profens (structure inset to Figure 5), but once again, all of the expected structural features are present in the 1H NMR spectrum (Figure 5). Thus, the para-substituted aromatic/phenyl ring displays its characteristic AA′BB′ splitting at ∼7.55 ppm while the aromatic protons from the indole moiety resonate between 6.6 and 7.0 ppm. In addition, the resonances for the two methyls (3.73 and 2.2 ppm) and the methylene group (3.64 ppm) are also clearly present (see Figure 5). Mass Spectra. Mass spectra were obtained for all four test compounds. Thus, the first eluting peak showed a molecular ion for naproxen at m/z 229.0843 (calcd 229.0865, C14H13O3) corresponding to the [M - D]-. This spectrum also contains peaks at m/z 185 and 276 which correspond to the loss of CO2 and the adduction of deuterated formic acid, respectively. Likewise the second eluting peak showed a molecular ion [M - D]- for flurbiprofen at m/z 243.0794 (calcd 243.0821, C15H12O2F), and also peaks at m/z 199 and 290. As with the naproxen spectrum, these ions correspond to the loss of CO2 and the adduction of deuterated (14) Hirschfeld T. Anal. Chem. 1980, 52, 297A-3012 A
3926 Analytical Chemistry, Vol. 72, No. 16, August 15, 2000
CONCLUSIONS While this prototype was constructed as a “proof of concept” and was by no means fully optimized, the results nevertheless demonstrate that there are no insurmountable difficulties in assembling a complex multiply hyphenated chromatographic system from commercially available spectrometers. It is selfevident that the individual spectrometers vary considerably in their sensitivities, and such a system can only be as sensitive as its least sensitive spectrometer. In this case, the amount of sample required is probably dictated by the NMR spectrometer (although in this configuration the FT-IR is of comparable sensitivity). Whatever the limitations of this device, it has still proved possible to obtain UV, IR, NMR, and MS spectra (together with atomic composition based on accurate mass determination) following a reversed-phase chromatographic separation. With modest refinement and optimization such a “total organic analysis device” might enable the analysis of complex mixtures for both confirmation of identity and structure determination (should this represent a costeffective approach to the solution of this type of problem). In addition, it is clear that we have by no means exhausted the possibilities for hyphenation and other types of spectrometers (e.g., fluorescence, circular dichroism, etc.) could also be included. However, while the current work clearly demonstrates that such multiple hyphenation is technically feasible, it is still the case, as pointed out by Hirschfeld many years ago,14 that merely because something is possible does not mean that it represents a good solution to a problem. Thus, having demonstrated that this type of system can be constructed, future work will therefore be directed toward applications to real samples rather than model mixtures of standards. Received for review February 23, 2000. Accepted June 2, 2000. AC000204Y
