Dynamic Biochemical Information Recovery in Spontaneous Human

As part of an ongoing study into the molecular basis of infertility following SCI, we investigated the time dependence of the reactions of SF from six...
0 downloads 0 Views 1MB Size
Download PDF
Anal. Chem. 2009, 81, 288–295

Dynamic Biochemical Information Recovery in Spontaneous Human Seminal Fluid Reactions via 1 H NMR Kinetic Statistical Total Correlation Spectroscopy Anthony D. Maher,*,† Olivier Cloarec,†,‡ Prasad Patki,§ Michael Craggs,| Elaine Holmes,† John C. Lindon,† and Jeremy K. Nicholson*,† Department of Biomolecular Medicine, Division of Surgery, Oncology, Reproductive Biology and Anaesthetics (SORA), Faculty of Medicine, Imperial College London, South Kensington, SW7 2AZ, United Kingdom, School of Biological Sciences, Royal Holloway University of London, Egham, Surrey, TW20 0EX, United Kingdom, and Department of Neuro-urology, and Division of Surgery and Interventional Sciences, University College London and Spinal Research Centre, Royal National Orthopaedic Hospital, Stanmore, Middlesex, HA7 4LP, United Kingdom Human seminal fluid (HSF) is a complex mixture of reacting glandular metabolite and protein secretions that provides critical support functions in fertilization. We have employed 600-MHz 1H NMR spectroscopy to compare and contrast the temporal biochemical and biophysical changes in HSF from infertile men with spinal cord injury compared to age-matched controls. We have developed new approaches to data analysis and visualization to facilitate the interpretation of the results, including the first application of the recently published K-STOCSY concept to a biofluid, enhancing the extraction of information on biochemically related metabolites and assignment of resonances from the major seminal protein, semenogelin. Principal components analysis was also applied to evaluate the extent to which macromolecules influence the overall variation in the metabolic data set. The K-STOCSY concept was utilized further to determine the relationships between reaction rates and metabolite levels, revealing that choline, N-acetylglucosamine, and uridine are associated with higher peptidase activity. The novel approach adopted here has the potential to capture dynamic information in any complex mixture of reacting chemicals including other biofluids or cell extracts. Metabolic profiling of biofluids has gained increasing importance in the context of systems biology approaches to understanding disease and metabolic responses of living systems to pathophysiological states.1 Readily obtainable biofluids such as urine and plasma give end-point makers of biological status, and analytical methods for their metabolic analysis by NMR spectros* To whom correspondence should be addressed. E-mail: j.nicholson@ imperial.ac.uk; [email protected]. † Imperial College London. ‡ Royal Holloway University of London. § Department of Neuro-urology, Royal National Orthopaedic Hospital. | Division of Surgery and Interventional Sciences, University College London and Spinal Research Centre, Royal National Orthopaedic Hospital. (1) Nicholson, J. K.; Lindon, J. C.; Holmes, E. Xenobiotica 1999, 29, 1181– 1189.

288

Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

copy have been well-documented.2-4 Other biofluids and secretions have received comparatively less attention, although characteristic metabolic signatures that vary according to function and disease have been described.5 Previous studies have investigated the metabolic composition of boar6 and human7 seminal fluid (SF) by NMR spectroscopy and made a substantial assignment of its metabolic content8 and its dynamic reaction behavior, including artificial prostate/seminal vesicle fluid mixtures.9 There is increasing interest in application of high-resolution 1H NMR spectroscopy to human seminal fluid (HSF). For example, recent studies have attempted to noninvasively diagnose prostatic tumors10,11 and test the effects of an injectable contraceptive on the distribution of amino acids.12 However, the full exploitation of 1H NMR for extraction of clinically relevant information is limited because of the incompleteness of the assignments and lack of understanding of the details of the dynamic reactions and their effect on the spectral profiles. HSF dilutes and transports spermatozoa to the ovum for fertilization and provides metabolic support to the cells. It is a complex, dynamic mixture of amino acids, sugars, peptides, proteins, and minerals. Immediately following ejaculation, a series of reactions occur, initiated by semenogelin (Sg) I and II (2) Lindon, J. C.; Holmes, E.; Nicholson, J. K. FEBS J. 2007, 274, 1140–1151. (3) Lindon, J. C.; Holmes, E.; Nicholson, J. K. Curr. Opin. Mol. Ther. 2004, 6, 265–272. (4) Nicholson, J. K.; Foxall, P. J.; Spraul, M.; Farrant, R. D.; Lindon, J. C. Anal. Chem. 1995, 67, 793–811. (5) Nicholson, J. K.; Wilson, I. D. Prog. Nucl. Magn. Reson. Spec. 1989, 21, 449–501. (6) Kamp, G.; Lauterwein, J. Biochim. Biophys. Acta 1995, 1243, 101–109. (7) Spraul, M.; Nicholson, J. K.; Lynch, M. J.; Lindon, J. C. J. Pharm. Biomed. Anal. 1994, 12, 613–618. (8) Lynch, M. J.; Masters, J.; Pryor, J. P.; Lindon, J. C.; Spraul, M.; Foxall, P. J.; Nicholson, J. K. J. Pharm. Biomed. Anal. 1994, 12, 5–19. (9) Tomlins, A. M.; Foxall, P. J.; Lynch, M. J.; Parkinson, J.; Everett, J. R.; Nicholson, J. K. Biochim. Biophys. Acta 1998, 1379, 367–380. (10) Averna, T. A.; Kline, E. E.; Smith, A. Y.; Sillerud, L. O.; Kline, E. E.; Treat, E. G.; Averna, T. A.; Davis, M. S.; Smith, A. Y.; Sillerud, L. O. J. Urol. 2005, 173, 433–438. (11) Kline, E. E.; Treat, E. G.; Averna, T. A.; Davis, M. S.; Smith, A. Y.; Sillerud, L. O. J. Urol. 2006, 176, 2274–2279. (12) Chaudhury, K.; Sharma, U.; Jagannathan, N. R.; Guha, S. K. Contraception 2002, 66, 199–204. 10.1021/ac801993m CCC: $40.75  2009 American Chemical Society Published on Web 11/26/2008

aggregation to form a gelatinous mass, which is then cleaved by prostate-specific antigen (PSA, also called kallikrein-related peptidase 3) after 5-20 min, resulting in liquefication.13,14 Subsequent to this, further reactions over the next 6-8 h result in hyperactivation and capacitation of the spermatozoa.15 It has been shown that disruptions to the biochemical and biophysical reactions following ejaculation are a major cause of infertility in men.16,17 Spinal cord injury (SCI) affects up to 200 000 men each year in the United States. Among the multitude of factors affecting rehabilitation in these patients is compromised sexual health postinjury.18 While the exact infertility mechanism remains unclear, decreased sperm motility (but not sperm count), modulated by poorly characterized factors in seminal plasma, has been the definitive observation.19,20 The extent to which SCI influences these dynamic reactions of HSF is unknown, largely due to the lack of an established analytical technique for simultaneous observation of the time dependence of both metabolites and proteins in this fluid. There is increasing interest in statistical correlation analysis of NMR data, since resonance intensities in a series of spectra are directly proportional to the amount of a given substance, provided the experiment is acquired under standardized conditions. This has proven useful for resolution enhancement in the indirect dimension in 2D NMR21 and for complex mixture analysis.22 An approach developed in our laboratory, statistical total correlation spectroscopy (STOCSY),23 has proven successful as a means for identification of latent biomarkers24 and for probing xenometabolome signatures.25 Recent extensions of this concept have enabled virtual enhancement of chromatographic resolution in hyphenated LC-NMR26 and S-DOSY, in which statistical correlations within a set of diffusion-ordered NMR spectra were used to identify latent biomarkers in epidemiological subpopulations.27 Investigations of the intensity correlations occurring during (13) Jonsson, M.; Linse, S.; Frohm, B.; Lundwall, A.; Malm, J. Biochem. J. 2005, 387, 447–453. (14) Pampalakis, G.; Sotiropoulou, G. Biochim. Biophys. Acta 2007, 1776, 22– 31. (15) Marieb, E. N.; Hoehn, K. Human Anatomy and Physiology, 7th ed.; Benjamin/Cummings: Menlo Park, CA,2007. (16) Sharma, R. K.; Agarwal, A. Urology 1996, 48, 835–850. (17) Hafez, B.; Goff, L.; Hafez, S. Arch. Androl. 1997, 39, 173–195. (18) Patki, P.; Woodhouse, J.; Hamid, R.; Craggs, M.; Shah, J. J. Spinal Cord. Med. 2008, 31, 54–60. (19) Brackett, N. L.; Davi, R. C.; Padron, O. F.; Lynne, C. M. J. Urol. 1996, 155, 1632–1635. (20) Brackett, N. L.; Nash, M. S.; Lynne, C. M. Phys. Ther. 1996, 76, 1221– 1231. (21) Bruschweiler, R.; Zhang, F. J. Chem. Phys. 2004, 120, 5253–5260. (22) Eads, C. D.; Noda, I. J. Am. Chem. Soc. 2002, 124, 1111–1118. (23) Cloarec, O.; Dumas, M.-E.; Craig, A.; Barton, R. H.; Trygg, J.; Hudson, J.; Blancher, C.; Gauguier, D.; Lindon, J. C.; Holmes, E.; Nicholson, J. Anal. Chem. 2005, 77, 1282–1289. (24) Holmes, E.; Cloarec, O.; Nicholson, J. K. J. Proteome Res. 2006, 5, 1313– 1320. (25) Holmes, E.; Leng Loo, R.; Cloarec, O.; Coen, M.; Tang, H.; Maibaum, E.; Bruce, S. J.; Chan, Q.; Elliott, P.; Stamler, J.; Wilson, I. D.; Lindon, J. C.; Nicholson, J. K. Anal. Chem. 2007, 79, 2629–2640. (26) Cloarec, O.; Campbell, A.; Tseng, L. H.; Braumann, U.; Spraul, M.; Scarfe, G. B.; Weaver, R.; Nicholson, J. K. Anal. Chem. 2007, 79, 3304–3311. (27) Smith, L. M.; Maher, A. D.; Cloarec, O.; Rantalainen, M.; Tang, H.; Elliott, P.; Stamler, J.; Lindon, J. C.; Holmes, E.; Nicholson, J. K. Anal. Chem. 2007, 79, 5682–5689.

reactions of complex mixtures have also been achieved using kinetic-STOCSY.28 The aim of the present work was to develop methods to analyze the 1H NMR observable postliquefication reactions in HSF (i.e., those that accompany capacitation) for assignment of the dynamic regions of the spectra, with a view to gaining insight into the mechanisms that result in infertility in patients with SCI and to identify potential biological or biophysical markers. MATERIALS AND METHODS Materials. Chemicals were purchased from Sigma (St. Louis, MO). D2O (99.9%) was from Goss Scientific Instruments Ltd. (Essex, UK). Sample Collection and Handling. After obtaining informed consent (and with ethical approval, NHS Research and Ethics Committee ref No. 05/Q0506/17), HSF was collected as part of an ongoing study based at the Royal National Orthopedic Hospital, Stanmore UK, by vibroejaculation from four individuals with SCI above thoracic vertebra 10 (T10) or by self-stimulation from three age-matched fertile (known to have fathered a child within 12 months) controls. Semen was snap frozen in liquid nitrogen within 5 min of ejaculation and later stored at -40 °C. Sample Preparation. For NMR analysis, samples were thawed at room temperature for 30 min and then centrifuged at 16000g for 5 min to separate the cells from the rest of the fluid. The supernatant was then diluted 2:1 in D2O containing 50 mg/ mL sodium trimethylsilyl[2,2,3,3-2H4]propionate. The 550-µL sample was transferred to a 5-mm NMR tube and inserted into the spectrometer. After allowing for temperature equilibration (to 300 K), the sample was shimmed and data were acquired after ∼60 min post-thaw. Experiments were acquired on a Bruker Avance-II NMR spectrometer equipped with a CryoProbe operating at 600.22-MHz 1H frequency. 1D NMR. Time course 1H NMR spectra were acquired with a pulse sequence of the form d1 - π/2x - t1 - π/2x - tm - π/2x - acquire, where π/2x represents a 90° hard pulse along the x-axis, d1 is a relaxation delay (4 s), t1 is a short delay (4 µs), and tm is a mixing time (10 ms). The resonance of H2O (at approximately δ4.7) was suppressed by selective irradiation during d1 and tm. For each spectrum in a time course, 16 transients were collected into 128k data points. Spectra were acquired at intervals of ∼10 min. Prior to Fourier transformation, the free-induction decay was multiplied by an exponential line-broadening factor of 1 Hz and zero filled by a factor of 1. Principal Components Analysis (PCA) and STOCSY. PCA models and STOCSY were achieved using in-house software (MetaSpectra, O. Cloarec, Imperial College London). NMR spectral data were imported into Matlab (Mathworks, Natick, MA) at full resolution (i.e., no bucketing was used) and without normalization. Time course data sets were arranged as n × k matrices (i.e., n time points and k variables). STOCSY analysis was achieved by calculating a vector of correlations (of length k) from a selected variable (corresponding to a peak of interest) to all other variables in the spectrum. To facilitate display, the square (28) Johnson, C. H.; Athersuch, T. J.; Wilson, I. D.; Iddon, L.; Meng, X.; Stachulski, A. V.; Lindon, J. C.; Nicholson, J. K. Anal. Chem. 2008, 80, 4886–4895.

Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

289

Figure 1. The 600-MHz 1H NMR spectra from HSF from a fertile volunteer. (A) Spectrum at 90 min postejaculation with key metabolites assigned. Key: lac ) lactate, spe ) spermine/spermidine, and uri ) uridine. Other amino acids have their standard abbreviations. Unassigned protein signals have been indicated. (B, C) Stacked plot of NMR spectra, color-coded by time postejaculation. p1 and p2 indicate the two different types of proteins visible in the envelope between δ0.5 and δ1.0.

of the correlation vector was projected as a color onto a plot of the covariance (as has been previously described23). RESULTS Time Course 1D 1H NMR Spectra from Human Seminal Fluid. The aliphatic region of a 1D 1H NMR spectrum of a typical HSF from a healthy volunteer 90 min postejaculation is shown in Figure 1A. Resonances from a number of low molecular weight metabolites can immediately be identified from previously assigned chemical shift and J-coupling patterns,8 such as choline, glycerophosphocholine (GPC), citrate, fructose, uridine, and lactate and resonances from amino acids such as alanine, valine, and isoleucine. Also observable in this spectrum are the characteristic broad features from proteins, but assignment of these in 1D spectra is hampered due to homogeneous broadening. The Sg are the most abundant proteins in human seminal plasma, at 19 mg/mL (∼0.3 mM),29 while PSA and albumin are found at levels of ∼0.6 mg/mL (∼0.018 mM),30 so it is reasonable to assume that some of the peaks visible in these spectra correspond to resonances from these proteins. When handled under standardized conditions, most biofluids used in metabonomics (e.g., urine and plasma) are stable with respect to their metabolic profiles within the time period required (29) Yoshida, K.; Yamasaki, T.; Yoshiike, M.; Takano, S.; Sato, I.; Iwamoto, T. J. Androl. 2003, 24, 878–884. (30) Alexandrino, A. P.; Rodrigues, M. A.; Matsuo, T. J. Urol. 2004, 171, 2230– 2232.

290

Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

for measurement.31-33 SF is different in that natural peptidase activity within the sample causes changes on the minute time scale. The most obvious manifestation of this in the 1H NMR spectrum is the increase in intensity of resonances from amino acids, with a concomitant decrease in signals from proteins. This is illustrated in Figure 1B, showing an expanded region of the spectrum in Figure 1A with spectra acquired over ∼5 h, colored according to the time post-thaw, with the red spectrum the final time point. This plot shows the two overlapping doublets from lactate and threonine near δ1.33. At any given time point, these 1D spectra can be ambiguous with respect to the assignment of these resonances to their respective metabolites, but when analyzed over time, it is clear that only two of the four peaks are increasing. This time-dependence information, combined with a basic knowledge of the physiology of the sample (e.g., peptidase activity) can then be used to assign these peaks to threonine (thr) and the others to lactate (lac). In Figure 1C, the spectral region between δ0.7 and δ1.0 has been expanded. This shows the narrow resonances from amino acids such as valine and isoleucine increasing, while resonances in the protein envelope are diminished (labeled p1). Interestingly, there are specific chemical shifts in the broader spectral regions (labeled p2) that are stable over time, suggesting different types of proteins are visible in these spectra and respond differently to peptidase activity. STOCSY works on the basic premise that, in a set of spectra of variable composition, the intensities of the resonances from a given molecule in NMR spectra will always have a fixed propor-

Figure 2. K-STOCSY analysis of time course NMR data from HSF, driven from the valine resonance at δ1.05. (A) 1H chemical shift region δ0.5-δ4.5 showing high correlations to other amino acid peaks and anticorrelations to protein signals. (B) 1H chemical shift region δ5.0-δ8.0. Amino acid signals are labeled with their standard abbreviations.

tionality, provided experimental acquisition parameters are the same. As such, this type of time course data is ideally suited to STOCSY analysis for exploration of its molecular components. The results of STOCSY “driven” from the valine doublet at δ1.04 in the set of time-course NMR spectra from HSF from a fertile volunteer are shown in Figure 2A (showing the aliphatic region of the spectra) and Figure 2B (the aromatic region). High correlations are observed to many resonances in the spectra, corresponding to signals from amino acids as they are released, and anticorrelated regions, corresponding to protein signals degraded by peptidase activity. Interestingly, there is not a uniform distribution of amino acids (from the 20 “standard” found in mammalian proteins) being released over time. For example, no tryptophan or cysteine was observed as increasing in this sample. This is consistent with the peptidase activity in this sample being due to PSA activity on SgI and SgII proteins predominantly, which

have combined distributions of tryptophan and cysteine of 0.28 and 0.48%, respectively.34,35 Concatenation of Time Course Data from Several Samples. As part of an ongoing study into the molecular basis of infertility following SCI, we investigated the time dependence of the reactions of SF from six samples, two from healthy volunteers and four from men with SCI of T10 and above. The small number of samples was due to constraints associated with the time taken for each experiment and that some of the samples were volume limited. In all six samples, peptidase activity was observed in the form of increasing intensities of resonances from amino acids. The STOCSY approach described above was extended by concatenating the time-series data matrices from each sample. This introduced additional variation across the data set due to initial concentration differences between samples, providing a more powerful means to extract structural information. We investigated Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

291

Figure 3. Concatenated K-STOCSY constructed from NMR spectra from all time points from six HSF samples ∼90-300 min postejaculation. (A) Driven from the “decreasing” peak at δ0.86. (B) Driven from the “static” peak at δ0.79.

the structural correlations in the 1D 1H NMR-visible protein resonances in these spectra by “driving” STOCSYs from the region δ0.5 and δ1.0 in increments of δ0.02. The results suggested there were two distinct types of proteins visible in these spectra. The results from STOCSY “driven” from δ0.86 and δ0.79 are representative of these and presented in Figure 3. When driven from δ0.86 (Figure 3 A), STOCSY revealed a correlation pattern across the spectra that was consistent with the decreasing spectral regions described above (as anticorrelated to the increasing valine signal), and this is consistent with assignment of these peaks to Sg I and II, the substrates for PSA.36 Note that the relatively narrow peaks from amino acids are not anticorrelated because of the increased variation across the data set introduced by concatenating data from several samples. STOCSY driven from δ0.79 revealed a different correlation pattern, including very broad features δ9.5. It is interesting to compare the correlations in the region near δ2.05, known to correspond to NHCOCH3 groups of N-acetyl sugars on glycoproteins.37,38 In Figure 3A, a correlated,

(31) Lauridsen, M.; Hansen, S. H.; Jaroszewski, J. W.; Cornett, C. Anal. Chem. 2007, 79, 1181–1186. (32) Maher, A. D.; Zirah, S. F.; Holmes, E.; Nicholson, J. K. Anal. Chem. 2007, 79, 5204–5211. (33) Teahan, O.; Gamble, S.; Holmes, E.; Waxman, J.; Nicholson, J. K.; Bevan, C.; Keun, H. C. Anal. Chem. 2006, 78, 4307–4318. (34) Malm, J.; Hellman, J.; Magnusson, H.; Laurell, C. B.; Lilja, H. Eur. J. Biochem. 1996, 238, 48–53. (35) Ulvsback, M.; Lazure, C.; Lilja, H.; Spurr, N. K.; Rao, V. V.; Loffler, C.; Hansmann, I.; Lundwall, A. J. Biol. Chem. 1992, 267, 18080–18084. (36) Lilja, H. World J. Urol 1993, 11, 188–191. (37) Bell, J. D.; Brown, J. C.; Nicholson, J. K.; Sadler, P. J. FEBS Lett. 1987, 215, 311–315. (38) Grootveld, M.; Claxson, A. W.; Chander, C. L.; Haycock, P.; Blake, D. R.; Hawkes, G. E. FEBS Lett. 1993, 322, 266–276.

Table 1. Resonance Assignments for Macromolecular Structures in 1H NMR Data from HSF chemical shift (δ)

assignment

0.79 0.84-0.88 0.91-0.92 1.12-1.15 1.17-1.19 1.21-1.23 1.25-1.30 1.39-1.42 1.65-1.7 2.05 P2 2.98-3.00 4.37-4.40 4.47-4.50 4.57-4.60 6.81-6.83 7.35-7.4 7.50-7.60

P2 Sg Sg P2 P2 Sg P2 Sg Sg Sg Sg P2 P2 Sg P2 P2

narrow resonance at δ2.06 was observed while in Figure 3B this was a broader resonance centered at δ2.05. Close inspection of the spectra confirmed that there were two different peaks identifiable in this region, implying there are different patterns of glycosylation for these two proteins identifiable by NMR. By integrating the results from the time dependence of these peaks and observed correlation patterns, we have assigned the correlation pattern observed in Figure 3A to Sg (I and II). The correlation pattern in Figure 3B is likely to be from a combination of other, nondegrading proteins such as PSA and albumin. Although the NMR data cannot conclusively determine if one or more proteins are responsible for these correlations, they can be considered to operate as a “functional unit” from a metabolic profiling point of view. We have denoted these as P2 in the figures and tables. These assignments are summarized in Table 1. Principal Components Analysis of Time Course Data from Human Seminal Fluid from Spinal Cord Injured Patients and Fertile Controls. Overall trends and variation in the data set were further probed using PCA. An important consideration in PCA is the type of scaling used, which is usually mean centering (MC) or unit-variance (UV) scaling. We constructed PCA models from the entire concatenated data set with both types of scaling and found that MC models were dominated by the very intense choline and citrate resonances (data not shown). The results from a UV-scaled PCA are presented in Figure 4. The scores plot (Figure 4A) has been color-coded according to the class of sample, with blue representing SF from SCI patients, and red fertile controls. Although no clear separation of the two groups was observed, it is interesting to note that the interindividual differences exceed those due to peptidase activity, as evidenced by tight clustering of scores from each individual in PC1 v PC2, which explained 60 and 19% of the variation in the data set, respectively. The PC loadings are plotted in Figure 3B and C, multiplied by the total standard deviation, and colored by the square of the same PC to aid visualization. Figure 3B shows that the PC1 loadings were dominated by the spectral regions identified above as originating from P2, while Figure 3C shows the Sg regions contributed the second highest source of variation, reinforcing the observation that two classes of protein can be observed in

Figure 4. UV-scaled PCA model of time course 1H NMR data of HSF from men with (blue) and without (red) spinal cord injury. (A) Scores plots for PC1 v PC2 (% variance explained in parentheses); (B, C) product of PC loadings with standard deviation of the entire data set (SD), color-coded by the square of the PC.

the 1D 1H NMR spectra. Note that variation in the data set as a result of peptidase activity was not observed until PC5, indicating its low contribution to the overall variation in the data set (for both MC and UV PCA, data not shown). Reaction Rate Correlations with Metabolic Signatures. The analysis was extended by investigating the relationship between the rates of reactions and amounts of metabolites. This was initially achieved by integrating the relatively isolated tyr Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

293

Figure 5. Reaction rate-driven STOCSY analysis of HSF. (A) Intensity of the resonance from tyrosine at δ6.90. Blue ) SCI patients; red ) fertile controls. The unbroken lines represent least-squares regressions of 2° polynomials. Asterisks have been used to denote the data from the SCI patient with highest rate tyrosine increase. (B) Bar graph showing the rate (min-1) of tyrosine release at 90 min postejaculation calculated from the 2° polynomials fit to the data in (A). (C) STOCSY “driven” from the rate of tyrosine release at 90 min postejaculation, showing correlations to fructose, choline, and Sg. Inset: spectral region corresponding to N-acetylglucosamine. (D) 1H NMR spectra from the region containing olefinic Hs from uridine. Blue ) SCI; red ) control. Note that the spectrum denoted by “*” is from the same sample denoted “*” in (A).

doublet at δ6.90 for each time course. The intensities over time are plotted in Figure 5A, where blue represents SCI patients and red the controls. It was observed that one of the samples from an SCI patient had a much higher rate of increase in tyr compared to the other three (represented by * in the figure). The unbroken lines drawn through each time course are 2° polynomials fitted to each data set, which were used to calculate the rate of reaction at 90 min postejaculation, plotted in Figure 5B. This rate was converted into a column vector of 6 values, which was then used as the “driver peak” for STOCSY to a data matrix constructed from the first spectrum from each time course. This revealed that the rate of increase of tyrosine was highly correlated to spectral regions corresponding to Sg, choline, fructose, uridine, and N-acetylglucosamine (Figure 5C and inset). To ensure this was not an artifact, this analysis was repeated by measuring the rate of change of other metabolites (such as the phenylalanine resonance at δ7.43) and after applying different types of normalization to the data (such as probabilistic quotient normalization, not shown). In each case, the choline, uridine, and N-acetylglucosamine were consistently highly correlated with reaction rate. Visual inspection also revealed that uridine levels were very low or unobservable in three of the four SF samples from SCI patients (blue traces, Figure 5D). Interestingly, the SCI patient with high uridine (represented by the blue asterisks in this figure) was the same patient with high rate of release of tyrosine observed in Figure 5A. 294

Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

DISCUSSION NMR spectroscopy is a widely used tool for elucidation and verification of chemical structures,39 this being one of the principal benefits of its application to metabolic profiling; unknowns can be often be identified from chemical shift and J-coupling patterns. Urine and blood plasma from healthy patients are more or less static with respect to metabolite concentration, so typically unknown resonances in such complex mixtures may be assigned by conventional 2D NMR experiments (COSY, TOCSY, HSQC, etc.) or by exploiting some physical property of the unknown molecule, such as its diffusion coefficient.40 In this paper. we have identified the spectral regions that continue to change 90 min postejaculation in HSF and show that these correspond to amino acids being released from Sg by peptidase activity. Additionally, we were able to combine within-sample (time course) and between-sample variation with STOCSY to assign specific spectral regions to macromolecular structures, in this case the resonances corresponding to Sg. While a previous report has published 1D NMR spectra of fragments of Sg,13 we are unaware of any that have attempted to assign the protein either purified or in HSF. With the increasing use of 1H NMR for investigating HSF, this information will be especially valuable for metabolic profiling studies that seek to identify biomarkers, since the recovery of biological information depends not just on how but also when the sample is spectroscopically measured. As we have shown, STOCSY is well suited to structural assignment in DOSY data because the peak intensity depends on

the Stokes’ radius-dependent translational diffusion coefficient, a property of the whole molecule.27 In the time course data sets (Figures 1 and 2), the proportions of the intensities of any given molecule will still remain fixed, and therefore, K-STOCSY can be used to detect correlations to molecules with similar reaction rates. This type of analysis becomes more powerful as another level of variation is introduced into the data set by adding the time course data from other samples, as shown in Figure 3, facilitating the assignment of the chemical shifts for macromolecules. Such information facilitates progression toward a complete assignment of the 1H NMR metabolic profile for HSF and will enhance the information obtainable from metabonomic studies investigating clinical states such as infertility. Usually in metabolic profiling studies, the term “time course” refers to samples collected from an individual at different times over the course of a study, while in this paper we are referring to spectra acquired from one sample at different time points. However, our application of these samples by PCA can still be considered analogous to metabolic trajectory analysis as has been applied to define treatment-specific profiles in toxicity studies.41-43 In the results presented here, there was no clear difference between SCI and controls with respect to their metabolic trajectories as defined by PC scores, using either MC or UV scaling. However, an interesting result was obtained from the UV-scaled PCA model, using a new approach to visualization of the loadings as presented in Figure 4B and C. Here we were able to identify two classes of macromolecules (proteins) that described the two major sources of variation within the data set. Care must be taken when interpreting this result, because just as one molecule of high intensity (such as choline) can dominate the PC loadings in a MC-scaled model, so one molecule (in this case a protein) can dominate the loadings in a UV-scaled model, due to the number of variables it contributes to in the NMR data. Thus, it does not imply that this protein has the highest variance

in the data set but rather that it occupies the most “chemical shift bandwidth”. Although the identity of this protein cannot be conclusively determined by these methods, the STOCSY analysis implies that it/they behave as a “functional unit” within NMR data; that is, they are biochemically related to each other enough to be considered one entity for metabonomics studies. Although no clear differences in terms of peptidase activity were observed between SCI and controls, the findings presented in Figure 5 raise several interesting hypotheses. We have found that there is a relationship between the rate of reaction and levels of Sg, choline, and N-acetylglucosamine. Choline is known to derive from the seminal vesicle (choline is produced from GPC within the first 60 min),8,9 as is Sg, implying that contributions from this vesicle determine the overall rate of these reactions, rather than the (prostate-derived) PSA. The observation that the only SCI sample that had high peptidase activity also had high uridine levels is also interesting. The function of uridine in HSF remains unclear, but has been shown to influence sperm velocity parameters and motility.44 Further recent work in our laboratory has revealed a strong association between seminal uridine levels and fertility in SCI.45 The application of K-STOCSY is of course not limited to spontaneously reacting mixturessindeed reactions induced by specific enzyme addition to human plasma have shown to be informative in assigning NMR-visible macromolecules in this biofluid,37 and elevated enzyme levels in disease states induce reactions that may be observed by this approach.46 The methods developed here would also be an efficient means of capturing the dynamic responses and potentially even the related changes in protein conformation and ligand binding or unfolding.

(39) Elyashberg, M. E.; Williams, A. J.; Martin, G. E. Prog. Nucl. Magn. Reson. Spectrosc. 2008, 53, 1–104. (40) Johnson, C. S. Prog. Nucl. Magn. Reson. Spectrosc. 1999, 34, 203–256. (41) Beckwith-Hall, B. M.; Nicholson, J. K.; Nicholls, A. W.; Foxall, P. J.; Lindon, J. C.; Connor, S. C.; Abdi, M.; Connelly, J.; Holmes, E. Chem. Res. Toxicol. 1998, 11, 260–272. (42) Bollard, M. E.; Holmes, E.; Lindon, J. C.; Mitchell, S. C.; Branstetter, D.; Zhang, W.; Nicholson, J. K. Anal. Biochem. 2001, 295, 194–202. (43) Keun, H. C.; Ebbels, T. M. D.; Bollard, M. E.; Beckonert, O.; Antti, H.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Chem. Res. Toxicol. 2004, 17, 579–587.

Received for review September 19, 2008. Accepted October 28, 2008.

ACKNOWLEDGMENT A.D.M. acknowledges funding from the EU Framework 6 MolPAGE project (LSHG-512066).

AC801993M (44) Niemeyer, T.; Dietz, C.; Fairbanks, L.; Schroeder-Printzen, I.; Henkel, R.; Loeffler, M. Nucleosides Nucleotides Nucleic Acids 2006, 25, 1215–1219. (45) Maher, A. D.; Patki, P.; Lindon, J. C.; Want, E. J.; Holmes, E.; Craggs, M.; Nicholson, J. K. Clin. Chem., in press. (46) Anthony, M. L.; Beddell, C. R.; Lindon, J. C.; Nicholson, J. K. J. Pharm. Biomed. Anal. 1993, 11, 897–902.

Analytical Chemistry, Vol. 81, No. 1, January 1, 2009

295