Use of an Atmospheric Solids Analysis Probe (ASAP) for High Throughput Screening of Biological Fluids: Preliminary Applications on Urine and Bile Marian Twohig,† John P. Shockcor,† Ian D. Wilson,‡ Jeremy K. Nicholson,§ and Robert S. Plumb*,†,§ Pharmaceutical Business Operations, Waters Corporation, Milford, Massachusetts, AstraZeneca, Clinical Pharmacology, Drug Metabolism and Pharmacokinetics, Alderley Park, Macclesfield, Cheshire SK10 4TG, U.K., and Division of Oncology and Surgery, Imperial College, London, SW7 Received February 8, 2010
A hybrid quadrupole orthogonal time-of-flight mass spectrometer (QToF) equipped with a solids analysis probe (atmospheric solids analysis probe-mass spectrometry (ASAP-MS)) has been applied to the high throughput qualitative analysis of bile (rat and dog) and urine (rat) samples. The metabolic profiles generated by ASAP-MS was less comprehensive than that provided by liquid chromatography (LC) or gas chromatography-mass spectrometry (GC-MS) metabonomic profiling, though simple types of sample preparation were found to increase the range of ions detected for bile (a complex, multicompartment sample type). While unsuited to biomarker discovery, ASAP-MS of these biofluids generated sufficiently complex metabolic fingerprints to enable them to be distinguished from each other using multivariate statistical methods such as principal components analysis (PCA). This ability to provide an effective means of sample classification suggests possible diagnostic applications. Keywords: Metabolic profiling • metabonomics • mass spectrometry • biofluid screening • urine • bile.
Introduction Personalized medicine strategies, delivered via the application of “omics” of various sorts (genomics, proteomics, metabonomics, etc.) are believed by many to hold the key to the future of drug development and human health care.1 The ability to determine, and understand, the individual biochemical profiles of patients (at the gene, protein, or metabolite level) may well allow the selection of the best combinations of drugs, at optimal doses, to be administered to a patient thereby minimizing nonresponders and adverse events. Personalizing the treatment to the patient via an appropriate diagnostic test should also allow drugs that might only work on a subset of the diseased population to be developed and marketed with obvious benefits to human health. Arguably, such biochemical profiling at the population level requires a rapid, inexpensive, information-rich, analytical technology that can be easily deployed in a clinical setting. Metabonomics, as shown by a number of studies where, for example, direct profiling of biofluids such as urine or plasma by nuclear magnetic resonance spectroscopy (NMR), has already provided significant insights into human disease (e.g., refs 2–4). Such studies, which have further evolved into the concept of pharmaco-metabonomics,5,6 demonstrate that metabonomics is well suited to addressing these challenges. Genomics by comparison does not take into account the effects * Author for correspondence. E-mail:
[email protected]. † Waters Corporation. ‡ AstraZeneca. § Imperial College, London.
3590 Journal of Proteome Research 2010, 9, 3590–3597 Published on Web 05/07/2010
of diet or environmental conditions, while proteomics is still comparatively expensive and time-consuming. Metabolic profiling in order to perform metabonomics was initially performed via proton NMR spectroscopy by Nicholson et al.7 However, more recently mass spectrometry, in the form of liquid chromatography-mass spectrometry (LC-MS), has become a useful alternative to NMR and, by providing complementary data, provides greater coverage of the metabolome,8,9 allowing the identification of a wider range of biomarkers. LCMS is a widely available technology and this will result in its widespread use for metabonomic research applications. However, the requirement for chromatographic separations to
Figure 1. Atmospheric solid analysis probe assembly. 10.1021/pr100120g
2010 American Chemical Society
research articles
High Throughput Screening of Biological Fluids
The advent of direct MS technologies such as desorption electrospray ionization (DESI) and direct analysis in real time (DART) has allowed the direct analysis of samples without the need for an LC separation. While these technologies are not as informative as LC-MS, or especially UPLC-MS, and hence would not lend themselves to biomarker discovery, they may well prove to be suitable for the rapid analysis of samples to be mapped against a validated model. Here we present the use of an atmospheric solids analysis probe (ASAP) as described by McEwan et al.11 for the high throughput metabonomic screening of biological fluids such as urine and bile from rats and dogs.
Experimental Section Figure 2. ASAP source.
maximize metabolome coverage often results in analysis times in the region of 15-20 min, although the application of sub 2 µm particle LC-MS (UPLC-MS) has allowed run times to be reduced to just 5-10 min or less.10 LC-MS also brings the requirement for an experienced analytical chemist and solvents, calibration, and knowledge to run the system. This is not a challenge when operating in a diagnostic laboratory; however, if MS is to be applied in a field-based diagnostics approach then a simpler technology is required that can be applied without the need for highly trained analysts, etc.
Chemicals. Methanol, acetonitrile, formic acid, and ammonium formate were obtained from Sigma-Aldrich (St. Louis, MO, USA). Distilled water 18 MOhm was produced in house using a Milli-Q system, Millipore (Billerica, MA, USA). Animal Samples. Control bile samples (24 h) were obtained, as described previously,12 via bile duct cannulation, from 24 male Sprague-Dawley rats, with six samples of control dog bile obtained from two dogs (three samples/dog on three different occasions). Quality control (QC) samples, comprised of a mixture of each of the bile samples analyzed (e.g., see refs 13 and 14) were prepared by careful mixing of an equal volume (20 µL) from each of the bile samples to create a pool sample that contained a representative mixture of all of the bile
Figure 3. Rat urine spectrum from ASAP analysis. Journal of Proteome Research • Vol. 9, No. 7, 2010 3591
research articles
Twohig et al.
Figure 4. Extracted ion pseudo chromatograms from rat urine analysis by ASAP.
samples tested. This QC mixture was then treated as per the normal bile samples. Rat urine samples (24 h) were also obtained from the bile cannulated rats. All of the samples were stored frozen at -20 °C prior to analysis by MS. Mass Spectrometry Conditions. Mass spectrometry was performed on a Waters Xevo QToF hybrid quadrupole orthogonal time-of-flight mass spectrometer. The instrument was operated in combined ESI/APCI mode (ESCI). This enabled the acquisition of analyte data in APCI mode and reference data in ESI mode. Ionization was performed in ESCI in positive ion mode for the urine samples and negative ion mode for the bile acids using a capillary voltage (ESI) of 3.0 kV with a corona current (APCI) of 5 µA. The source conditions were optimized to use a cone voltage of 40 V, and aperture 1 setting of 15 V, desolvation temperature of 100-450 °C, desolvation gas 500 L/h, and a source temperature of 120 °C. The MS data were acquired over a range of 100-1000 m/z with a scan duration of 0.5 s, using leucine enkephanin as the lock reference (ESI). Sample Loading. The samples were analyzed in a random order, with biological QC samples intermixed at regular intervals between them (every six samples) as well as water blanks. The samples were coated onto the outside of a sealed glass capillary by simply dipping the capillary into the biofluid and allowing the bulk of the excess to drain off. Any remaining excess was then removed using a stream of nitrogen and the sample was taken immediately for analysis. A fresh capillary was used for each analysis. Data Analysis. The resulting low collision energy MS data from the QToF was processed using MarkerLynx XS. The data was treated, automatically, by the summation of all of the spectra produced in the sample. As there was no separation 3592
Journal of Proteome Research • Vol. 9, No. 7, 2010
technology involved, it was not necessary to time align the data set for retention time drift. The spectra/intensity from each sample was normalized for intensity against the sum of the peak intensities and then multiplied by 10 000 to give a final response for each peak in the spectrum. The data from each sample was then combined to give an overall data set. Zeros were inserted in the response table for each sample which did not detect an ion that was present in the master data set. A full description of this methodology can be found in U.S. Patent 7,418.352.B2.
Results and Discussion The ASAP process for liquids such as the urine and bile examined here was both rapid and technically undemanding. The operator simply dips the end of the sealed glass capillary into the sample, thereby coating the outside with liquid, followed by removal of the excess. The capillary is then attached to the MS probe and the whole assembly (Figure 1) is inserted into the mass spectrometer; a new glass capillary is used for each sample to minimize the chances of carry over and cross contamination. Once the probe has been inserted into the mass spectrometer the sample was desorbed from the probe by a heated stream of gas, which vaporized any volatile compounds. The sample is ionized in the gas phase by an APCI-like process as described by McEwan et al.,11 Figure 2. This process can be performed rapidly such that MS data can be acquired at rate of one sample analyzed every 40-60 s (even though currently sample changing is performed manually). In this study, we observed no carry over between samples as evidenced by the absence of signals observed for the water blanks. The acquired
research articles
High Throughput Screening of Biological Fluids
Figure 5. ASAP pseudo chromatogram and spectra from bile analysis. Table 1. Table of Masses Identified As Contributing to the Observed Variance in the Data from Bile Analysisa mass (m/z)
p[1]P
p(corr)[1]P
229.153 86.0959 90.064 263.241 265.256 271.27 116.066 118.087 120.072 287.242 295.268 297.288 130.067 299.299 132.103 133.107 132.304 313.277 314.282 319.266 147.114 331.286 332.29 337.28 339.295 341.309 342.315
0.0703617 0.113362 0.0714874 -0.130702 -0.087938 -0.061647 0.0824036 0.151155 0.143226 -0.073602 -0.064527 -0.061282 0.115261 -0.080289 0.286665 0.0825892 0.0706576 -0.170618 -0.076245 -0.073797 0.154719 -0.162064 -0.068899 -0.077394 -0.081274 -0.152918 -0.071995
0.779966 0.898051 0.924479 -0.90099 -0.9417 -0.89616 0.915434 0.876469 0.801986 -0.9512 -0.97194 -0.9819 0.826178 -0.96876 0.925362 0.918857 0.758948 -0.98413 -0.98172 -0.96333 0.790233 -0.9816 -0.97549 -0.92417 -0.9594 -0.98743 -0.98863
a The values p[1]P relate to the loadings value and p(corr)[1]P to the correlation value from the OPLS-DA analysis.
MS data takes on the appearance of a broad peak as different ions are vaporized from the probe, as shown in the specific examples given below. Urine Analysis. Analysis of the urine samples showed that it was possible to detect many of the ions previously reported in LC/MS metabolic profiling studies. Thus, Figure 3 shows the full scan MS obtained from the positive ion ASAP analysis of urine, where a total of some 7000 ions were observed including, for example, diagnostic ions from hippurate (m/z )180), pantothenic acid (m/z ) 220), kynurenic acid (m/z ) 190), xanthenic acid (m/z ) 206), phenylpuruvate (m/z ) 165) and indican (indoxyl sulfate) (m/z ) 212). The extracted ion MS traces for these ions are displayed in Figure 4. This result pro-
Table 2. Positive Ion ASAP-QToF-MS of Bile Acid Standards bile acid standard
MX- iona
glycholic acid taurocholic acid cholic acid Na glycochenodeoxycholate dehydrocholic acid lithocholic acid sodium glycholate deoxycholic acid chenodeoxycholic acid Na taurochenodeoxycholate Na taurodeoxycholate
407 406 407 575 401 375 575 391 391 390 436
second ion
third ion
446 433
464 452
fragment ion 255 447 421 fragment ion 255 437 437 436
a The MX- ion is the negative ion nominal mass of detected analyte, with the second and third ions related to fragment or aduct ions observed with each bile acid.
Table 3. Marker Ions Derived From The PCA Analysis of the Bile Samples Following Addition of Methanol, where p[1] and p[2] Relate to the Coefficient of Variance from the PCA Loadings Plot mass (m/z)
p[1]
p[2]
101.0212 143.032 161.0399 255.2302 279.2279 283.2599 281.2448 304.2354 303.226 390.2989 406.2911 436.2953
-0.0892013 -0.0825043 -0.149759 -0.144704 -0.156761 -0.159877 -0.145005 -0.0777542 -0.164246 -0.0756819 -0.0849576 -0.130356
-0.034492 -0.0402257 0.0368924 0.0823256 0.129895 0.0699177 0.0777991 0.0476693 0.104272 0.0440647 0.109549 0.109581
vides some confidence that, while not providing as comprehensive a metabolite profile as, for example, UPLC-MS, the technique is capable of analyzing the urine samples derived from metabonomic studies and detecting known diagnostic ions, and may thus potentially be capable of being used in a sample screening/diagnostic role. Multivariate statistical analysis of these data for sample classification is discussed below. Bile Analysis. Bile analysis was initially undertaken on untreated biofluid samples as for urine. The positive ion MS data displayed in Figure 5 shows the result obtained for a typical dog bile sample. The two inset MS spectra, A and B, Journal of Proteome Research • Vol. 9, No. 7, 2010 3593
research articles
Twohig et al.
Figure 6. Comparison of dog bile and rat bile spectra from ASAP analysis.
were derived from the beginning and end of the vaporization process, respectively. The data presented clearly show that the MS spectrum obtained from the beginning of data acquisition was different from that at the end. The tail end of the peak shows compounds with a much higher m/z value than those early in the sample vaporization, and this is most likely due to the different boiling points and vapor pressures of the various compounds contained in the sample. These MS data showed a significant number of ions (570 in negative ion mode), but interestingly ions for the bile acids, which might be expected to be significant contributors to the profile, were absent. As seen with the urine samples, the 3594
Journal of Proteome Research • Vol. 9, No. 7, 2010
number of ions detected for bile was significantly less, by a factor of 10, than that normally observed by UPLC-MS of bile.12 In addition, a simple comparison of the ASAP analyses with those previously undertaken by UPLC-MS12 confirmed that the ions observed were quite different, tending to be more fragment-ion related rather than for the expected intact bile acids (the major ions detected for bile are listed in Table 1). This difference may in part have been be due to the mode of ionization, which, as we have indicated, is more APCI-like in nature rather than electrospray but may also have been due to bile acid ionization having been suppressed by their incorporation into micelles in the bile. When standards of the
High Throughput Screening of Biological Fluids
research articles
Figure 7. PCA (PC1 vs PC2) analysis of entire data set including bioanalysis.
Figure 8. (a) PCA (PC1 vs PC2) analysis of rat urine (blue), rat bile (red), and dog bile (green); (b) PCA (PC1 vs PC2) analysis of rat bile (purple) and dog bile (black) only. Journal of Proteome Research • Vol. 9, No. 7, 2010 3595
research articles
Twohig et al.
Figure 9. PCA analysis of rat and dog bile following addition of methanol prior to ASAP analysis.
individual bile acids were dissolved in water/methanol 1:1 v/v and analyzed using the ASAP the bile acids were readily detected providing some support for this notion. The ions produced by the ASAP analysis of the bile acid standards were not the MH- ions but pseudo molecular ions such as, for example, that observed for taurocholic acid (m/z 452) which can be rationalized as resulting from the formation of the sodium adduct (ASAP-MS data for these bile acid standards are given in Table 2). The addition of methanol to bile samples provides a means to disrupt the micelles and should thus facilitate their detection using the ASAP probe. The rat and dog bile samples in the present study were therefore diluted with an equal volume of methanol and then reanalyzed by ASAP QToF MS whereupon signals relating to bile acids (probably taurocholic, deoxycholic acid, and taurodeoxycholic acid) were detected (see Table 3). Representative MS spectra of the rat and dog bile following methanol dilution and analysis using the ASAP are displayed in Figure 6. Following this type of treatment some 7500 ions were seen in the bile samples. In the dog bile spectrum ions can clearly be seen corresponding to taurocholic acid at m/z ) 452, deoxycholic acid m/z ) 391, and Na taurodeoxycholate at m/z ) 436. In the rat bile data, the taurocholic acid ions at m/z ) 452 and 406, and the Na taurodeoxycholate at m/z ) 436 were clearly observed. Metabonomic Analysis. There were approximately 7000 MS features in the combined data set for these (untreated) biofluids which, as might be expected from the method of analysis, was somewhatlowerthanthatpreviouslyreportedusingUPLC-MS.12,14,15 However, this still represents a large number of ions, and certainly many more than could be examined manually, and therefore chemometric analysis was performed. Thus, the data from the ASAP-QToF MS were subjected to multivariate statistical analysis, initially using the unsupervised approach of principal components analysis (PCA). The PCA scores plot displayed in Figure 7 shows the result (PC1 vs PC2) for all the samples in the original data set, including rat urine and unprocessed rat and dog bile, together with QC samples, blanks, etc. From this analysis, the QC samples clustered tightly together, as did the blanks, thus indicating that the intrasample sample repeatability was acceptable providing some confidence in the quality of the data.13,14 The next step of the data analysis was 3596
Journal of Proteome Research • Vol. 9, No. 7, 2010
to remove the QC samples from the data set for PCA (the resulting scores plot is shown in Figure 8a). Here the dog bile samples were clearly separated from the rat bile samples, as was the case when UPLC-MS was performed on the same samples.12 Perhaps not unexpectedly, given the major differences in metabolite composition between urine and bile, this analysis showed that both of the bile data sets were clearly separated from those resulting from the rat urine samples. However, while not unexpected the separation of the urine from the bile samples, and the rat and dog bile samples from each other, demonstrates that sufficient information was obtained to allow the discrimination of the data set. This gives confidence that the data generated by the ASAP probe technique, while not suitable for obtaining full metabolite profiles, nevertheless do contain sufficient information to allow the two groups to be separated suggesting potential, for example, for diagnostic applications. When the data analysis was restricted to just the bile samples (Figure 8b), the dog bile samples were seen to cluster tightly while the rat bile samples showed a greater degree of variability, similar to that seen when these samples were analyzed using UPLC/MS.12 The statistical analysis of the bile data obtained after methanol treatment also showed excellent clustering of the various sample types with excellent grouping of the QCs as before. The PCA scores plot shown in Figure 9 shows the results for PC1 and PC2, with the QCs and blanks removed. As can be seen rat and dog bile samples are, as seen without methanol treatment, clearly separated from each other with the dog bile samples once again clustering tightly together while those for the rats showed greater variability. The ions shown as responsible for the observed statistical separation of these rat and dog bile samples were then determined by means of an OPLS-DA S-Plot. In this type of statistical treatment, the ions at the extremities of the S-Plot are those that contributed most to the observed clustering (data not shown) with the ions m/z ) 406, m/z ) 436 (probably representing taurocholic acid ion and the Na taurodeoxycholate ion) identified as key contributing ions to the statistical analysis.
Discussion These preliminary studies on urine and bile samples using the ASAP probe show that a simple and direct analysis of the biological samples using ASAP-MS can be performed enabling
research articles
High Throughput Screening of Biological Fluids the rapid and technically undemanding metabonomic screening of biological samples. Thus, the removal of the hyphenation of the MS to either LC or GC systems means that this approach can be performed without the need for a highly trained analytical chemist and complex separation step, perhaps making it a possible platform for in-clinic analysis. However, as the results for bile show, it may still be necessary to perform some sample manipulation in order to detect particular compounds/metabolites. Thus, while the ASAP technique, as would indeed be expected for any direct MS analysis methodology, revealed fewer features for both urine and bile than one based on, for example, LC-MS successful discrimination between the various sample types was possible via a variety of “diagnostic” ions. The limitations of this device for generating comprehensive metabolite profiles clearly indicated that, at least in its present form ASAP should not be used in studies where the aim is investigative sample profiling/biomarker discovery. However, where biomarkers of disease or toxicity, etc. have already been identified using a more comprehensive approach such as, for example, UPLC-MS then it may well prove to be possible to build a rapid and cost-effective method based on them that can be deployed in a clinical setting or for population screening in, for example, epidemiological investigations where the aim is diagnosis/screening. However, in such clinical applications it will clearly be necessary to have fully defined and annotated the sets of marker ions identified as discriminating between healthy and diseased subjects for them to be accepted as biomarkers.
Conclusions The ASAP-MS process enables rapid analysis of biological samples such as bile and urine. Although the data generated are less comprehensive than that which can be obtained using
LC or GC-MS, making it unsuited to biomarker discovery, it does enable effective sample classification. Simple types of sample preparation can extend the range of ions detected for complex, multicompartment, sample types such as bile.
References (1) Holmes, E.; Wilson, I.; Nicholson, J. Cell 2008, 134, 714–717. (2) Lindon, J. C.; Holmes, J. C.; Bollard, M. E.; Stanley, E. G.; Nicholson, J. K. Biomarkers 2004, 9, 1–31. (3) Brindle, J. T.; Antti, H.; Holmes, E.; Tranter, G.; Nicholson, J. K.; Bethell, H. W.; Clarke, S.; Schofield, P. M.; McKilligin, E.; Mosedale, D. E.; Grainger, D. J. Nat Med. 2002, (12), 1439–44. (4) Mowatt, G.; Zhu, S.; Kilonzo, M.; Boachie, C.; Fraser, C.; Griffiths, T.; N’dow, J.; Nabi, G.; Cook, J.; Vale, L. Health Technol. Assess. 2010, 14 (4), 1–356. (5) Clayton, T. A.; Lindon, J. C.; Cloarec, O.; Antti, H.; Charuel, C.; Hanton, G.; Provost, J. P.; Le Net, J. L.; Baker, D.; Walley, R. J.; Everett, J. R.; Nicholson, J. K. Nature 2006, 440, 1073–1077. (6) Clayton, T. A.; Baker, D.; Lindon, J. C.; Everett, J. R.; Nicholson, J. K. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 14728–14733. (7) Wilson, I. D.; Nicholson, J. K. Prog NMR Spectrosc. 1989, 21, 449– 501. (8) Theodoridis, G.; Gika, H. G.; Wilson, I. D. TrAC - Trends Anal. Chem. 2008, 27, 251–260. (9) Wu, Z.; Huang, Z.; Lehman, R.; Zhao, C.; Xu, G. Chromatographia 2009, 69, S23–S32. (10) Plumb, R. S.; Granger, J. H.; Stumpf, C. L.; Johnson, K. A.; Smith, B. W.; Gaulitz, S.; Wilson, I. D.; Castro-Perez, J. Analyst 2005, 130, 844–849. (11) McEwan, C. N.; McKay, R. G.; Larsen, B. S. Anal. Chem. 2005, 77, 7826–7831. (12) Plumb, R.S.; Rainville, P. D.; Potts, W. B.; Johnson, K. A.; Gika, E.; Wilson, I. D. J Proteome Res. 2009, 8, 2495–2500. (13) Gika, H. G.; Theodoridis, G. A.; Wingate, J. E.; Wilson, I. D. J. Proteome Res. 2007, 6, 3291–3303. (14) Gika, H. G.; Macpherson, E.; Theodoridis, G.; Wilson, I. D. J Chromatogr., B 2008, 871, 299–305. (15) Wilson, I. D.; Nicholson, J. K.; Castro-Perez, J.; Granger, J. H.; Johnson, K.; Smith, B. W.; Plumb, R. J. Proteome Res. 2005, 4, 591–598.
PR100120G
Journal of Proteome Research • Vol. 9, No. 7, 2010 3597
