Anal Bioanal Chem (2010) 396:1451–1463 DOI 10.1007/s00216-009-3338-z
ORIGINAL PAPER
Identification of biochemical changes in lactovegetarian urine using 1H NMR spectroscopy and pattern recognition Jingjing Xu & Shuyu Yang & Shuhui Cai & Jiyang Dong & Xuejun Li & Zhong Chen
Received: 26 August 2009 / Revised: 4 November 2009 / Accepted: 21 November 2009 / Published online: 17 December 2009 # Springer-Verlag 2009
Abstract A vegetarian diet has been demonstrated to have a profound influence on human metabolism as well as to aid the prevention of several chronic diseases relative to an omnivorous diet. However, there have been no systematic metabolomic studies on all of the biochemical changes induced in human subjects by long-term vegetarianism. In this study, 1H NMR spectroscopy in combination with multivariate statistical analysis was applied to explore the variability in the metabolic urinary profiles of healthy populations from four groups: lactovegetarian male (VEGMALE), lactovegetarian female (VEGFEMALE), omnivorous male (OMNMALE), and omnivorous female (OMNFEMALE). Differences in metabolic profiles were examined in relation to diet and gender by principal component analysis (PCA) and spectral integrals. It was found that the most influential low molecular weight metabolites responsible for the differences between the diet groups were N-acetyl glycoprotein (NAG), succinate, citrate, trimethylamine-N-oxide (TMAO), taurine, glycine, hippurate, phenylalanine, methylhistidine and formate, whereas for the differences in gender groups the most discriminatory metabolites were NAG, succinate, creatinine, arginine, TMAO, taurine, hippurate, mannitol, phenylalaElectronic supplementary material The online version of this article (doi:10.1007/s00216-009-3338-z) contains supplementary material, which is available to authorized users. J. Xu : S. Cai (*) : J. Dong : Z. Chen Department of Physics, Fujian Key Laboratory of Plasma and Magnetic Resonance, State Key Laboratory of Physical Chemistry of Solid Surfaces, Xiamen University, Xiamen 361005, China e-mail:
[email protected] S. Yang : X. Li (*) The First Hospital of Xiamen Affiliated to the Xiamen University, Xiamen 361001, China e-mail:
[email protected]
nine, and methylhistidine. The results from the PCA of all four groups indicated that diet plays a greater role in influencing metabolite differences than gender. As an exploration, this work shows the potential of metabolomics when applied to nutritional and physiological studies, and it will aid further studies. Keywords NMR . Metabolomics . Lactovegetarian . Gender . Principal component analysis . Metabolic profiles Abbreviations VEGMALE VEGFEMALE OMNMALE OMNFEMALE PCA NAG DMA TMAO ATP TCA
Lactovegetarian male Lactovegetarian female Omnivorous male Omnivorous female Principal component analysis N-Acetyl glycoprotein Dimethylamine Trimethylamine-N-oxide Adenosine triphosphate Tricarboxylic acid
Introduction Vegetarian diets have been associated with reduced risks for several chronic diseases and conditions, including obesity [1], coronary artery disease [2–4], diabetes [5, 6], and some types of cancer [7, 8]. Vegetarian diets are characterized as the consumption of plant foods such as grains, legumes, fruits, vegetables, nuts and seeds, whereas a “lactovegetarian” diet includes dairy products but excludes eggs, and a “lacto-ovovegetarian” diet includes both eggs and dairy products. Compared to an omnivore diet, a vegetarian diet
1452
has been linked to less weight gain [9], lower serum LDL and HDL cholesterols and triglycerides [10], and increased body leanness [11], although vegetarians have been reported to be at risk of vitamin B12 [12] and iron deficiency [13]. However, these findings do not tend to generate information on the mechanisms associated with such dietary benefits. Furthermore, diets, together with smoking, stress, exercise, etc., are considered external physiological factors, whereas gender, together with age, hormonal status, etc., are considered internal physiological factors. Any alteration in these factors can disrupt homeostasis, resulting in perturbations in the levels of endogenous biochemicals that are involved in the key metabolic processes in organisms [14]. The compositions of biofluids alter in response to such changes in tissue biochemistry. Therefore, we may hope to obtain valuable information regarding the underlying molecular mechanisms by monitoring these perturbations. The quantitative measurement of the dynamic multiparametric metabolic response of a living system to pathophysiological stimuli or genetic modification is termed “metabolomics” [15]. Among the analytical techniques used in metabolomics, 1H-NMR spectroscopy has been shown to be one of the most important, as it can detect many endogenous metabolites rapidly and reproducibly without derivatization or separation [16]. Several groups have studied the influence of physiological variation on the urine metabolic patterns of healthy human subjects using NMR-based metabolomic techniques [17–21]. NMR-based metabolomic techniques have also been used in diet-related studies. For example, Rezzi and coworkers studied the metabolic phenotypes correlated with two dietary preferences, namely “chocolate desire” and “chocolate indifference,” and showed that dietary preference could influence basal metabolic and gut microbiome activities [22]. Holmes and coworkers investigated metabolic phenotype variations across and within four human populations from China, Japan, the UK and the USA, and associated metabolic diversity with diet and blood pressure [23]. Stella and coworkers characterized the short-term effects of three diets, “vegetarian,” “low meat” and “high meat,” on the metabotype signatures of human participants [24]. Solanky and coworkers identified the biochemical effects of a diet rich in soy isoflavones on the 1H-NMR spectral profiles of urine collected from premenopausal women before and after the consumption of soy or miso [25]. van Dorsten and coworkers compared the effects of black and green tea consumption on human metabolism [26]. In this study, healthy male or female participants following long-term lactovegetarian or omnivorous diets are investigated using high-resolution 1H NMR spectroscopy in order to generate information on subtle diet- or gender-induced changes in urinary metabotypes. Character-
J. Xu et al.
istic metabolites are determined by principal component analysis (PCA) and spectral peak intensity analysis. The results validate the applicability of NMR-based metabolomics to dietary nutrition, and help to explore and interpret the advantages and disadvantages of vegetarian intake.
Methods and materials Sample collection This study was conducted in accordance with the medical ethical principles of the First Hospital of Xiamen and Xiamen Diabetes Institute. The respective local ethics committees approved the protocol before the commencement of the study, and all subjects gave written informed consent. Both lactovegetarian and omnivorous subjects were included in the study. Forty-one married male volunteers were military officers recruited from Xiamen Garrison, while 40 female volunteers were military spouses from the same source. These male and female individuals could come from different families. They normally consumed their habitual diets, containing foods of both plant and animal origins, at home during the study. These omnivorous individuals aged from 23 to 55 were considered the control group. Forty-two male together with 38 female lactovegetarians from Minnan Buddhist College and Nan Putuo Temple (both located in Xiamen) who had consumed a diet that excluded all animal products for more than five years were enrolled. These lactovegetarians were aged from 18 to 40, lived in a self-contained community, and usually ate meals from a communal kitchen where the food was prepared strictly according to lactovegetarian principles. The types of food eaten daily, along with estimated amounts of each, were: 480 g/day of vegetables, 350 g/day of fruit (including juice), 360 g/day of bean curd or other soy foods, 220 g/day of milk, 240 g/day of edible mushrooms, and 350 g/day (female) or 600 g/day (male) of rice. Women subjects who were menstruating were excluded from this exploratory study in order to avoid confounding effects on metabolic profiles. To be eligible to join the experiment, all of the subjects had to complete a questionnaire in which they confirmed their continued good health and that they did not take any regular medication. In addition, their clinical chemistries and hematological parameters had to be within acceptable limits. Exclusion criteria included diabetes, abnormal plasma glucose levels, hypertension, acute illness within two weeks before sample collection, and treatment in the preceding 2–3 months with any drug known to have potential hepatotoxic or nephrotoxic side effects. Subjects were requested to abstain from alcohol consumption for
Identification of biochemical changes in lactovegetarian urine using 1H NMR spectroscopy and pattern recognition
24 h before sampling. Overnight urine samples were collected from all subjects and centrifuged (3000 rpm, 5 min at 4 °C) to remove the particulate contaminants, and then stored in a freezer at −20 °C until NMR measurements were performed. Sample preparation and 1H NMR spectroscopy For each urine sample, a volume of 0.4 mL was mixed with 0.2 mL phosphate buffer solution (0.2 M Na2HPO4/0.2 M NaH2PO4, pH 7.4, 100% D2O) to minimize the variation in pH. 0.3 mM DSS (2,2-dimethyl-2-silapentane-5-sulfonic acid) was used as an internal reference standard at δ 0.0. The mixture was pipetted into a 5 mm NMR tube. The 1H NMR measurements were performed on a Varian (Palo Alto, CA, USA) NMR System 500 MHz spectrometer equipped with a triple-resonance probe. The temperature was set to 298 K and 90° pulse lengths were calibrated individually for each sample. A conventional presaturation pulse sequence for solvent suppression based on the one-dimensional version of the NOESY pulse sequence known as NOESYPR (nuclear Overhauser effect spectroscopy with presaturation: delay90°-t1-90°-tm-90°-acquisition) was used [27]. t1 was set to 2 μs, the mixing time tm was 120 ms, and weak irradiation was applied to suppress the signal from water during the recycling delay of 2 s and the mixing time. A total of 256 scans with a spectral width of 5 kHz were collected for all NMR spectra. All of the signals were zero-filled to 16 k before Fourier transformation (FT). The assignment of endogenous urinary metabolites in the 1H NMR urine spectra was performed by referring to published data [28]. Quantification of the urinary metabolites was based on the integrated peak areas of some selected signals that were normalized with respect to the total integrated area of each spectrum. Statistical analysis was performed with Statistica V5.5 (StatSoft Inc., Tulsa, OK, USA). Values were expressed as mean±SD and compared using ANOVA with Tukey analysis. Preprocessing of NMR spectra and principal component analysis (PCA) The collected NMR spectra were phased and baselinecorrected manually using the software MestRe-C 4.8 (http://www.mestrec.com). All spectra were also peakaligned manually to overcome peak shift problem [29]. The chemical shift regions δ 4.6–6.0 (water and urea resonances), δ 0.0–0.2 and δ 1.6–1.8 (DSS resonance) were excluded prior to PCA. Each spectrum was then binned into 201 segments with intervals of δ 0.04 across the range δ 0.2–9.8. To account for overall variations in sample concentration, each spectrum was normalized to its total integrated area.
1453
PCA was carried out on all NMR data using housewritten software written in C++. The PCA algorithm can be found in the literature [30]. The software was validated by comparing its results with those from commercial software. PCA is a projection method for exploring classificatory differences and highlighting explanatory metabolites. After PCA, the data are projected onto a few latent variables (or “principal components”) that describe the maximum variation within the data. Each principal component (PC) is a new variable created from a linear combination of original variables from a dataset with appropriate weighting coefficients. The first PC contains the largest proportion of the variance of the dataset, and the subsequent PCs contain smaller and smaller proportions of the variance. A score plot of the first two or three PCs best illustrates the inherent similarities and differences between the samples in response to various external stimuli. Each point on the score plot represents an individual sample. A loading plot reveals the contributions of various variables (integral segments) to the classification. Each point on the loading plot represents a single NMR spectral segment. Using the score and loading plots, the samples can be classified and the biochemical components responsible for this classification can be identified.
Results and discussion In the current study, we investigated the variability of the metabolic phenotypes of urine samples from healthy volunteers of both sexes following strict long-term lactovegetarian diets. To assess the effects of different diet constituents on human metabolism, the volunteers following an omnivorous diet were considered the control group. The dataset was divided into four classes: lactovegetarian male (VEGMALE), omnivorous male (OMNMALE), lactovegetarian female (VEGFEMALE), and omnivorous female (OMNFEMALE). First of all, the whole dataset from the four groups was imported for pattern recognition analysis in order to find the trends in the two effects (diet and gender). These two physiological factors are expected to compete with each other, and theoretically one of them will have a greater influence on the classification and clustering. The score plot in Fig. 1 shows that the four groups show distinct clustering based on the lactovegetarian and omnivorous diets, whereas the gender-related difference is obscured by the larger diet effect. 1
H NMR spectra of urine
Typical 1H-NMR spectra of urine from lactovegetarian and omnivorous subjects show a large number of isolated and
1454
J. Xu et al.
Fig. 1 PCA score plot showing typical effects of diet together with gender (inverted triangles, VEGMALE; circles, VEGFEMALE; diamonds, OMNMALE; squares, OMNFEMALE)
overlapping peaks, indicating that a large number of metabolites are excreted in urine due to human metabolism (see the “Electronic supplementary material,” Fig. S1). The spectra were assigned according to the literature [28]. The aliphatic regions of the spectra are dominated by peaks from creatinine (which gives rise to two intense peaks at δ 3.05 and 4.06), TMAO, glycine, citrate, succinate, dimethylamine (DMA) and taurine, whereas hippurate, methylhistidine, phenylalanine and formate generate large resonances that are visible in the aromatic region. Comparisons between the spectra from lactovegetarian and omnivorous subjects in both the male and the female groups demonstrate that the most marked differences associated with diet discrepancies, observable in the 1H NMR spectra, include decreased urinary concentrations of creatinine, taurine and TMAO in lactovegetarians. The spectral differences between the male and the female groups, on the other hand, were too subtle to be visible. Statistical classification was then applied to systematically inspect diet-related and gender-related effects on urinary composition. Analysis of dietary differences Four PCA models were constructed pair-wise based on the dataset. The VEGMALE–OMNMALE and VEGFEMALE–OMNFEMALE models were used to investigate diet effects on both male and female subjects, while the VEGMALE–VEGFEMALE and OMNMALE–OMNFEMALE models concerned the gender-induced metabolic discrepancy for vegetarian or omnivorous dietary intake. For each model, the first five PCs explain more than 80% of the spectral variance in the original spectra (see “Electronic supplementary material,” Table S1). To aid the rational recovery of metabolic information as well as the
optimal separation of the two different groups, the two PCs with the best separabilities were selected from among PC1– PC5 to use for the score plot of each model in the following discussion. Diet-dependent clustering of the lactovegetarian and omnivorous groups is visible in both male and female individuals (Fig. 2a). To interpret the score plot still further, a visual comparison of the spectra from two outlying lactovegetarian samples (marked as Nos. 24 and 41 in Fig. 2a) with spectra from others in the same group was conducted. It is obvious that these two outliers both have an unusually weak hippurate signal (δ 7.61, 7.82) relative to those seen in other lactovegetarian samples. Similarly, the two outliers of the OMNMALE groups (marked as Nos. 49 and 75 in Fig. 2a) were also compared to the spectra from others in the same group. The first outlier has very high citrate level (δ 2.54, 2.70), which can probably be attributed to personal-specific dietary preferences regarding vegetables or fruit, while the second outlier shows an unusual elevation in the aromatic region (δ 6.80–8.46). These four individuals were not excluded from the data analysis because these fluctuations were considered to be within the range of physiological variability of the metabolite profiles. The loading plot (Fig. 2b) represents a summary of the input variables and detects the most important variables for group separation. Variables (i.e., bins) containing similar information are grouped together in the loading plot, while those that departed from the origin show the contribution to inter- or intra-group discrepancy. The differentiation between the VEGFEMALE and OMNFEMALE groups is also shown in the PCA score plot (Fig. 2c). Improved separation of the two classes and more compact clustering of each class is observed relative
Identification of biochemical changes in lactovegetarian urine using 1H NMR spectroscopy and pattern recognition
1455
Fig. 2 Dietary differences highlighted by the PCA of urines from lactovegetarian and omnivorous people. a PCA score plot (PC2 vs. PC3) for the VEGMALE (inverted triangles) and OMNMALE (diamonds) groups; b PCA loading plot corresponding to score plot
(a); c PCA score plot (PC1 vs. PC3) for the VEGFEMALE (circles) and OMNFEMALE (squares) groups; d PCA loading plot corresponding to score plot (c)
to the VEGMALE–OMNMALE model. Continuous lactovegetarian dietary intervention shows a greater impact on the urinary metabolites of female subjects. The contributions of different variables to the classification are labeled in the loading plot (Fig. 2d). The possible characteristic metabolites obtained from the PCA models of VEGMALE–OMNMALE and VEGFEMALE–OMNFEMALE are listed in Table 1. In order to estimate the concentration variations of the metabolites, the differences in the means for the two models were calculated for each bin to reflect the mean variation of each bin (data not shown). Whether this difference is of statistical significance was assessed by ANOVA analysis. The results are shown in Table 1. The metabolites that change concentration in the VEGFEMALE–OMNFEMALE model are almost the same as those that do so in the VEGMALE–OMNMALE model, although citrate, and glucose do not show evident alterations in the VEGFEMALE–OMNFEMALE model.
A lower level of urinary taurine excretion in lactovegetarian subjects is observed upon visual spectral comparison and when studying the PCA loading plots. Taurine is a conditionally essential amino acid that plays an important role in many physiological functions in the human body. However, human beings have a limited ability to synthesize taurine and are probably dependent in part on dietary taurine, which is found exclusively in foods of animal origin. The vegetarian diet was low not only in taurine itself but also in substrates or in cofactors (e.g., methionine and cysteine) for taurine synthesis. The amount of urinary taurine in vegans was reported to be only 29% of the control value following chronic abstention from the consumption of any animal products [31, 32]. Although the clinical implications of a low dietary taurine intake and reduced urinary taurine levels in vegetarians remain unknown, the functional importance of taurine is growing increasingly evident. Taurine deficiency has been implicat-
1456 Table 1 Endogenous urinary metabolites responsible for the classification of omnivorous and lactovegetarian groups
Note: “↑” and “↓” denote higher or lower amounts relative to the omnivorous male or female group. * p
