NMR-based metabolomic studies reveal changes in biochemical profile of urine and plasma from rats fed with sweet potato fiber or sweet potato residue

Abstract

Dietary fiber has attracted more interest in recent years because many studies have uncovered its disease preventive and health-promoting features: containing blood cholesterol and/or glucose attenuation and reducing obesity risk. However, the health effects of sweet potato fiber (SF) and sweet potato residue (SR) and the knowledge of their exact mechanisms of action are still not fully understood. This study investigates the effect of SF and SR administration on rat metabolism. Rats were randomly assigned to one of three dietary groups of 11 rats each and given a basal diet containing 15% SF, 15% SR, or no supplemental fiber (control). The groups were observed for 30 days, and urine and plasma samples were analyzed by proton nuclear magnetic resonance. Compared with the rats in the control group, those administered the diet containing SF exhibited significantly increased plasma levels of lipid, lactate, and myo-inositol, as well as urine levels of acetate, citrulline, N-acetylglutamate, and p-hydroxyphenylacetate. SF significantly decreased the plasma levels of glutamine, glutamine/glutamate, lysine, phosphorylcholine/glycerolphosphocholine, tyrosine, and glucose, as well as the urine level of allantoin. Moreover, SR significantly increased the plasma levels of acetone, lipid, and lactate, as well as the urine levels of acetamide, alanine, citrulline, ethanol, lactate, methylamine, methylmalonate, N-acetylglutamate, and α-hydroxybutyrate, compared with the control group. SR significantly decreased the plasma levels of citrate, glutamate, glutamine, isoleucine, lysine, methionine, and glucose, as well as the urine level of isobutyrate. These results suggest that SF and SR supplementation have certain systemic metabolic processes in common, including lipid metabolism, glycogenolysis and glycolysis metabolism, energy metabolism, protein biosynthesis, and gut microbiota metabolism. This study demonstrates the potential for the routine use of metabolomics in nutritional studies to characterize metabolic effects and understand the influence of diet on animal metabotypes.

---

Introduction

Dietary fiber (DF) is defined as “edible parts of plants or analogous carbohydrates that are resistant to digestion and absorption in the human small intestine and are completely or partially fermented in the large intestine”.¹ Based on its solubility in water, DF can be classified into two major groups: soluble and insoluble. Soluble fibers include pectins, gums, and β-glucans, whereas insoluble fibers include cellulose, lignin, and hemicellulose.^2–4 Both soluble and insoluble fibers can serve as prebiotics for gut microbiota. The products of fiber fermentation involve short-chain fatty acids; namely, acetate, propionate, and butyrate.⁵ These fatty acids can be used as a source of energy for the host, and may affect normal gastrointestinal functions. Propionate can inhibit the activity of the enzyme 5-hydroxy-3-methylglutaryl-coenzyme A (HMG CoA) reductase, the limiting enzyme (EC 1.1.1.88 and EC 1.1.1.34; the enzyme commission designation is EC 1.1.1.34 for the NADPH-dependent enzyme, whereas 1.1.1.88 links to an NADH-dependent enzyme) of the mevalonate pathway (the metabolic pathway that produces cholesterol and other isoprenoids for cholesterol synthesis).⁶ Butyrate is the preferred energy source for the epithelial cells.⁷

DF has significant health implications in preventing the risks of constipation, obesity, cardiovascular disease, atherosclerosis, colon cancer, and diabetes.⁸ The physiological effects of DF are probably due to its physicochemical properties, such as water- and oil-holding capacities, absorption of organic molecules, bacterial degradation, cation-exchange capacity, and antioxidant activity.⁹

Sweet potato is an excellent source of natural health-promoting compounds. Approximately 85 million tons of sweet potatoes are produced annually; most of these come from China, which claims 77.5% of world production (FAO, 2008). In China, the main commercial use of sweet potatoes involves starch and starchy food production, which generates a large quantity of sweet potato residues (SRs). These residues are used to produce animal feed, or are discarded as waste materials. The SR waste creates problems related to disposal and potential severe pollution; it also represents a loss of valuable biomass and nutrients. Thus, the environmentally friendly conversion of SRs into useful products with higher value as by-products is urgently needed. Apart from starch, DF is the main component of SR. Generally, DF products are composed of cellulose, lignin, pectin, and hemicelluloses, which are present at 31.19, 16.85, 15.65, and 11.38 g per 100 g of dry matter, respectively.¹⁰ The relative monosaccharide contents of DF follow the order: glucose > uronic acid > galactose > arabinose > xylose > rhamnose > mannose.¹⁰ A previous study has suggested that sweet potato increases HDL-C and decreases LDL-C in humans, while moderately increasing serum glucose and cholesterol levels.¹¹ However, only a few studies have focused on the response of animal or human biological systems to sweet potato fiber (SF) and SR supplementation. In this study, we applied a nontargeted proton nuclear magnetic resonance (¹H NMR)-based metabolomic approach to investigate the global metabolic response to SF or SR supplementation in rats.

Metabolomics facilitates the understanding of global metabolite changes in animals or humans in response to changes in nutrition, genetics, environment, and gut microbiota.^12–14 This field has become increasingly important in understanding biological processes. The information obtained from metabolomic analysis can also provide a glimpse into the phenotypic state of an organism because the metabolic network is downstream of gene expression and protein synthesis. At present, metabolic profiles can be more comprehensively characterized using high throughput analytical tools such as ¹H NMR spectroscopy. NMR can provide the simultaneous quantitative measurement of numerous metabolites, allowing a more complete picture of the metabolic response to diet as well as novel insights into unrecognized mechanisms.¹⁵ Recently, metabolomics has been widely adopted in various biomedical, toxicological, and nutritional studies because any alteration in physiological status can disrupt homeostasis, resulting in perturbations in the levels of endogenous biochemicals involved in different key metabolic processes.¹² Thus, monitoring perturbations in biofluid composition can yield valuable information about molecular mechanisms and provide novel insights into changes in physiological status in biological systems.¹³ The combination of NMR and pattern recognition methods such as principle component analysis (PCA), projection to latent structure-discriminate analysis (PLS-DA), and orthogonal projection to latent structure-discriminate analysis (OPLS-DA) allows better visualization of changing endogenous biological profiles attributed to a physiological challenge or stimulus. These challenges or stimuli include disease processes, xenobiotic administrations, genetic modifications, or even nutritional interventions.

This approach can enhance the understanding of the metabolic status of SF or SR administration, provide clues on the relationship between metabolites and nutritionally influenced biochemical SF and SR mechanisms, and contribute baseline data for future metabolomic experiments on SF and SR supplementation. This approach is also potentially useful in investigating SF and SR metabolism and searching for further correlations between SF and SR administration and health or disease risk. Furthermore, this study can help define the effects of metabolic modifiers and refine nutritional requirements to enable better formulation of nutritional support for growth and health. By conducting an explorative metabolomic analysis using ¹H NMR spectroscopy and chemometrics, this study aims to investigate the effect of SF and SR administration on the urine and plasma compositions of rats.

Materials and methods

Animal experiments and sample collection

The experimental protocol used in this study was approved by the Animal Care and Use Committee of the Animal Nutrition Institute, Sichuan Agricultural University, and was carried out according to the National Research Council's Guide for the Care and Use of Laboratory Animals. A total of 33 eight-week-old female Sprague-Dawley rats weighing 170 g to 201 g were placed in individual metabolic cages and allowed to acclimatize for one week. The rats were randomly assigned to one of the three isonitrogenous and isocaloric dietary groups, which contained 11 rats per group, for 30 days. The experimental diets included a control diet without a supplemental fiber source and two diets with 15% SF (Shanxi Sciphar Hi-tech Industry Co., Ltd. Shanxi, China) or 15% SR (Zhenzhou Foods Processing Co., Ltd. Henan, China). The diets (Table S1†) were formulated to meet the nutrient recommendations of the AIN 39 for laboratory rodents.¹⁶ To ensure similar gross energy levels in all of the diets, corn starch and soybean oil were decreased in the SF and SR source diets. The body weight of each rat was recorded at the beginning and end of the trial. The daily feed intake of the rats was also recorded. The dosage in this study was selected based on a prior experiment.¹⁷ Contents of crude protein, soluble fiber, insoluble fiber, and total fiber are listed in Table 1. Feed and drinking water were allowed ad libitum. Temperature (22 °C to 25 °C), photoperiod (12 h light/12 h dark), and humidity (50% to 70%) were maintained throughout the study. Clinical observations were conducted during the experimental period, and bodyweights were determined once a week. Urine samples were collected into ice-cooled vessels containing 30 μL of sodium azide solution (1.0% w/v) from day 28 to day 29 of the treatment period (24 h). After anaesthesia with ether at the end of a 30 d treatment period, blood samples from the orbital venus plexus were collected (0900 a.m.) into Eppendorf tubes containing sodium heparin. Plasma samples were obtained by high-speed centrifugation at 3500 g for 10 min at 4 °C. All urine and plasma samples were stored at −80 °C until further testing.

Table 1 Contents of the fiber sources

Items	Sweet potato fiber	Sweet potato residue
Crude protein (%)	0	1.9
Soluble fiber (%)	100	9.8
Insoluble fiber (%)	0	20
Total fiber (%)	100	29.8

---

Sample preparation and NMR spectroscopy

A 550 μL aliquot of each urine sample was mixed with 55 μL of phosphate buffer (1.5 M NaH₂PO₄/K₂HPO₄, pH 7.4, 100% v/v D₂O) containing 5.0 mM 2,2-dimethyl-2-silapentane-5-sulfonate-d₆ (DSS) as a chemical shift reference (δ 0.00 ppm). After 10 min of centrifugation (12 000 rpm, 4 °C), the supernatant was pipetted into 5 mm NMR tubes for NMR analysis. Plasma samples were prepared by mixing 200 μL of plasma with 400 μL of saline solution containing 75% D₂O as a field frequency lock. Approximately 550 μL of each sample was then transferred into 5 mm NMR tubes after vortexing and centrifugation at 11 200 rpm for 10 min at 4 °C.

The proton NMR spectra of the urine and plasma samples were recorded at 300 K on a Bruker Avance II 600 MHz spectrometer (Bruker Biospin, Rheinstetten, Germany) operating at a ¹H frequency of 600.13 MHz and equipped with a broadband-observe probe. A standard water-suppressed one-dimensional NMR spectrum was obtained for urine using the first increment of the gradient-selected NOESY pulse sequence (recycle delay–90°–t₁–90°–t_m–90°–acquire data) with recycle delay of 2 s, t₁ of 3 μs, mixing time (t_m) of 100 ms, and 90° pulse length of 13.70 μs. A total of 128 transients were acquired in 49 178 data points using a spectral width of 9590 Hz and an acquisition time of 2.56 s. For plasma, a water-presaturated Carr–Purcell–Meiboom–Gill pulse sequence (recycle delay–90°–(τ–180°–τ)_n–acquisition) was employed to attenuate the NMR signals from macromolecules. A spin–spin relaxation delay (2nτ) of 76.8 ms and a spin–echo delay τ of 400 μs were used. Typically, the 90° pulse was set to 13.7 μs, and 32 transients were collected in 49 178 data points for each spectrum using a spectral width of 9590 Hz. Other acquisition parameters were the same as described above. Metabolite assignments were usually made by consideration of the chemical shifts, coupling constants, and relative intensities, as in previous reports.^18,19 Additional ¹H–¹H correlation and ¹H–¹H total correlation spectra were acquired for selected samples (data not shown).

Data processing and multivariate data analysis

Prior to Fourier transform, the free induction decays were multiplied by an exponential window function with a 1 Hz line-broadening factor. All NMR spectra were manually adjusted for possible phase and baseline distortions. The plasma spectral region δ 0.5 to δ 9.0 was integrated into regions with equal 0.002 ppm widths, and the urinary spectral region δ 0.5 to δ 9.5 was bucketed into regions with 0.01 ppm widths using Mestrenova 8.1.2 software (Mestrelab Research S.L., Spain). Plasma and urine chemical shifts were referenced to the peak of the L-lactate methyl proton at δ 1.33 and the DSS peak at δ 0.00, respectively. The ethanol signals originating from the blood collection process were carefully excluded together with the regions containing urea and H₂O signals to obtain only the signals from endogenous metabolite changes induced by SF and SR exposure. The urea signal was excluded to avoid any contributions of urea to intergroup differentiations in order to eliminate the effects of variation in urea signal caused by partial cross-solvent saturation via solvent-exchanging protons. Plasma urea was achieved by conventional chemistry measurements. This treatment can avoid any contributions of ethanol, urea, and H₂O to intergroup differentiations. The discarded regions in the plasma spectra included δ 4.19 to δ 5.23 and δ 5.40 to δ 6.80 for H₂O and urea, as well as δ 1.16 to δ 1.19 and δ 3.60 to δ 3.62 for ethanol. The discarded regions in the urine spectra included δ 4.50 to δ 5.00 for H₂O and δ 5.45 to δ 6.00 for urea. Subsequently, each integral region was normalized to the total sum of all integral regions for each spectrum before pattern recognition analysis.

Multivariate data analysis was performed on the normalized NMR datasets with the software package SIMCA-P+ (version 11.0, Umetrics, Sweden). PCA was carried out on the dataset to generate an overview of the sample distribution and observe possible outliers. PLS-DA and OPLS-DA were further performed with the unit-variance-scaled NMR data as the X matrix and class information as the Y matrix to identify the metabolites that significantly contribute to intergroup differentiation.²⁰ The quality of the model was described by the parameters R²X and Q², which represent the total explained variations for the X matrix and the model predictability, respectively. The models were certified via a seven-fold cross validation method and a permutation test.^21,22 A model was considered significant when the Q² value was significant (P < 0.05) through permutation. In order to facilitate the interpretation of results, the loadings that indicated altered metabolites after treatment were back-transformed in Excel (Microsoft, USA) and plotted with color-coded absolute coefficient values (|r|) of the variables in MATLAB (The Mathworks Inc.; Natwick, USA, version 7.1).²¹ The color-coded correlation coefficients indicate the significance of the metabolite contribution in predicting the response; those contributing the most to the prediction of the response (class) are shown in red, whereas those with slight or no association with the response are shown in blue. In this study, the correlation coefficient cutoff values depended on number of samples needed in each group to obtain statistical significance based on the discrimination significance of the Pearson's product-moment correlation coefficient at the level of P < 0.05.²¹

Results

¹H NMR spectra of urine and plasma samples

Fig. 1 shows typical ¹H NMR spectra of the urine samples taken from randomly selected rats in the SF, SR, and control groups. Fig. 2 shows the representative spectra of rat plasma from the SF, SR, and control groups. NMR signals were assigned to the ¹H resonances of specific metabolites (Table 2). A total of 45 metabolites were unambiguously assigned for urine. The urine sample spectra contained resonances from several amino acids and organic acids as well as glucose, allantoin, and choline. Tricarboxylic acid cycle metabolites such as succinate and citrate were also detected in the urine samples. The plasma samples mainly contained glucose, lactate, lipids, and a series of amino acids.

Fig. 1 Typical one-dimensional ¹H NMR spectra of urine metabolites obtained from the (A) control, (B) sweet potato fiber, and (C) sweet potato residue groups. For clarity, the δ 6.2–9.5 region was magnified 10 times, compared with the corresponding δ 0.5–6.2 region. Metabolite keys are given in Table 1.

Fig. 2 Representative 600 MHz ¹H NMR spectra of plasma metabolites obtained from the (A) control, (B) sweet potato fiber, and (C) sweet potato residue groups. For clarity, the δ 6.0–9.0 region was magnified 24 times, compared with the corresponding δ 0.5–6.0 region. Metabolite keys are given in Table 1.

Table 2 ¹H NMR data for metabolites in rat urine and plasma

Keys	Metabolites	Moieties	δ ^1H (ppm) and multiplicity	Samples^a
a U, urine; P, plasma; * LDL, low density lipoprotein; VLDL, low density lipoprotein; s, singlet; d, doublet; t, triplet; q, quartet; dd, doublet of doublets; m, multiplet.
1	Bile acids	CH3	0.62(m), 0.75(m)	U
2	Butyrate	CH3	0.9(t)	U
3	α-Hydroxybutyrate	CH3	0.94(t)	U
4	α-Hydroxy-iso-valerate	δCH3	0.97(d)	U
5	Isobutyrate	CH3	1.14(d)	U, P
6	Ethanol	CH3, CH2	1.19(t), 3.66(q)	U, P
7	Methylmalonate	CH3, CH	1.26(d), 3.76(m)	U
8	α-Hydroxy-n-valerate	CH3, γCH2	0.89(t), 1.31(m)	U
9	Lactate	αCH, βCH3	4.13(q), 1.33(d)	U, P
10	Alanine	αCH, βCH3	3.77(q), 1.48(d)	U, P
11	Citrulline	γCH2, βCH2	1.56(m), 1.82(m)	U
12	Acetate	CH3	1.92(s)	U, P
13	Acetamide	CH3	1.99(s)	U
14	N-Acetylglutamate	βCH2, γCH2, CH3	2.07(m), 1.88(m), 2.04(s)	U
15	Acetone	CH3	2.25(s)	U, P
16	Acetoacetate	CH3	2.3(s)	U
17	Succinate	CH2	2.41(s)	U
18	α-Ketoglutarate	βCH2, γCH2	2.45(t), 3.01(t)	U
19	Citrate	CH2	2.55(d), 2.68(d)	U, P
20	Methylamine	CH3	2.62(s)	U
21	Dimethylamine	CH3	2.73(s)	U
22	Trimethylamine	CH3	2.88(s)	U
23	Dimethylglycine	CH3	2.93(s)	U
24	Creatine	CH3, CH2	3.04(s), 3.93(s)	U, P
25	Creatinine	CH3, CH2	3.04(s), 4.05(s)	U, P
26	Ethanolamine	CH2	3.13(t)	U
27	Malonate	CH2	3.16(s)	U
28	Choline	OCH2, NCH2, N(CH3)3	4.07(t), 3.53(t), 3.20(s)	U, P
29	Taurine	–CH2–S, –CH2–NH2	3.26(t), 3.43(t)	U
30	TMAO	CH3	3.27(s)	U
31	Glycine	CH2	3.57(s)	U
32	Phenylacetyglycine	2,6-CH, 3,5-CH, 7-CH, 10-CH	7.31(t), 7.37(m), 7.42(m), 3.68(s)	U
33	Hippurate	CH2, 3,5-CH, 4-CH, 2,6-CH	3.97(d), 7.57(t), 7.65(t), 7.84(d)	U
34	N-Methylnicotinamide	CH3, 5-CH, 4-CH, 6-CH, CH2	4.44(s), 8.18(d), 8.89(d), 8.96(d), 9.26(s)	U
35	β-Glucose	1-CH, 2-CH, 3-CH, 4-CH, 5-CH, 6-CH	4.65(d), 3.25(dd), 3.49(t), 3.41(dd), 3.46(m), 3.73(dd), 3.90(dd)	U, P
36	α-Glucose	1-CH, 2-CH, 3-CH, 4-CH, 5-CH, 6-CH	5.24(d), 3.54(dd), 3.71(dd), 3.42(dd), 3.84(m), 3.78(m)	U, P
37	Allantoin	CH	5.40(s)	U, P
38	Urea	NH2	5.82(s)	U
39	Homogentisate	6-CH, 5-CH	6.7(d), 6.76(d)	U
40	p-Hydroxyphenylacetate	6-CH, 2-CH, 3,5-CH	3.6(s), 6.87(d), 7.15(d)	U
41	m-Hydroxyphenylacetate	6-CH, 4-CH, 3-CH	6.92(m), 7.04(d), 7.26(t)	U
42	Nicotinate	2,6-CH, 4-CH, 5-CH	8.62(d), 8.25(d), 7.5(dd)	U
43	4-Aminohippurate	CH2	7.71(d)	U
44	Trigonelline	2-CH, 4-CH, 6-CH, 5-CH, CH3	9.12(s), 8.85(m), 8.83(dd), 8.19(m), 4.44(s)	U
45	Formate	CH	8.46(s)	U
46	Unknown		8.54(s)	U
47	LDL*	CH3(CH2)n	0.87(m)	P
48	VLDL*	CH3CH2CH2C=	0.89(t)	P
49	Isoleucine	αCH, βCH, βCH3, γCH2, δCH3	3.68(d), 1.99(m), 1.01(d), 1.26(m), 1.47(m), 0.94(t)	P
50	Leucine	αCH, βCH2, γCH, δCH3	3.73(t), 1.72(m), 1.72(m), 0.96(d), 0.97(d)	P
51	Valine	αCH, βCH, γCH3	3.62(d), 2.28(m), 0.99(d), 1.04(d)	P
52	Propionate	CH3, CH2	1.08(t), 2.18(q)	P
53	3-Hydroxybutyrate	αCH2, βCH, γCH3	2.28(dd), 2.42(dd), 4.16(m), 1.20(d)	P
54	Lipids (triglycerids and fatty acids)	(CH2)n, CH2CH2CO, CH2C=C, CH2CO, C=CCH2C=C	1.28(m), 1.58(m), 2.01(m), 2.24(m), 2.76(m)	P
55	Lysine	αCH, βCH2, γCH2, εCH2	3.76(t), 1.91(m), 1.48(m), 1.72(m), 3.01(t)	P
56	N-Acetyl glycoprotein	CH3	2.04(s)	P
57	O-Acetyl glycoprotein	CH3	2.08(s)	P
58	Glutamate	αCH, βCH2, γCH2	3.75(m), 2.12(m), 2.35(m)	P
59	Methionine	αCH, βCH2, γCH2, S–CH3	3.87(t), 2.16(m), 2.65(t), 2.14(s)	P
60	Pyruvate	CH3	2.37(s)	P
61	Glutamine	αCH, βCH2, γCH2	3.78(m), 2.14(m), 2.45(m)	P
62	Glycerolphosphocholine	CH3, βCH2, αCH2	3.22(s), 3.69(t), 4.33(t)	P
63	Phosphorylcholine	N(CH3)3, OCH2, NCH2	3.22(s), 4.21(t), 3.61(t)	P
64	myo-Inositol	1,3-CH, 2-CH, 5-CH, 4,6-CH	3.60(dd), 4.06(t), 3.30(t), 3.63(t)	P
65	Threonine	αCH, βCH, γCH3	3.58(d), 4.24(m), 1.32(d)	P
66	Unsaturated lipids	=CH–CH2C=, –CH=CH–	5.19(m), 5.30(m)	P
67	Tyrosine	2,6-CH, 3,5-CH	7.20(dd), 6.91(d)	P
68	1-Methylhistidine	4-CH, 2-CH	7.05(s), 7.78(s)	P
69	Phenylalanine	2,6-CH, 3,5-CH, 4-CH	7.32(m), 7.42(m), 7.37(m)	P
70	3-Methylhistidine	4-CH, 2-CH	7.07(s), 7.67(s)	P

---

Multivariate data analysis of NMR data

PCA was initially conducted on the plasma spectral data. Two principal components were determined for the treatment groups; PC1 and PC2 explained 50.3% and 22.6% of the variables, respectively. The PCA results (Fig. 3A) demonstrated that differences between rats in the SF, SR, and control groups were not seen in their metabolic plasma profiles. Furthermore, the plasma metabolic changes in the rats from all groups were determined using OPLS-DA. The corresponding coefficient analyses showed that SF significantly increased the plasma levels of lipid, lactate, and myo-inositol, while SR significantly increased the plasma levels of acetone, lipid, and lactate, compared with the control group (P < 0.05). In contrast, SF significantly decreased the plasma levels of glutamine, glutamine/glutamate, lysine, phosphorylcholine/glycerolphosphocholine, tyrosine, α-glucose, and β-glucose, while SR significantly decreased the plasma levels of citrate, glutamate, glutamine, isoleucine, lysine, methionine, α-glucose, and β-glucose, compared with the control (P < 0.05, Fig. 4A and B and Table 3). SF significantly decreased the plasma levels of lysine, phosphorylcholine/glycerolphosphocholine, α-glucose, and β-glucose, compared with the SR group (P < 0.05, Fig. 4C and Table 3).

Fig. 3 PCA score plots (A, R²X = 0.912, Q² = 0.802) based on the ¹H NMR spectra of plasma metabolites obtained from the control (black squares), sweet potato fiber (green triangles), and sweet potato residue (red circles) groups, and PLS-DA score plots (B, R²X = 0.319, R²Y = 0.707, Q² = 0.193) based on the ¹H NMR spectra of the urine obtained from urinary metabolites of the control (black squares) and sweet potato residue (red circles) groups.

Fig. 4 OPLS-DA score plots (left panel) and the corresponding coefficient loading plots (right panel) of plasma metabolites derived from the control (black squares), sweet potato fiber (green triangles), and sweet potato residue (red circles) groups (A: R²X = 29.8%, Q² = 0.32; B: R²X = 30.6%, Q² = 0.017; C: R²X = 23.2%, Q² = 0.208). One control sample was excluded because it lay outside the Hotelling's T² ellipse on the score plot. The colour scale in the coefficient plot shows the significance of metabolite variation between the sweet potato fiber, sweet potato residue and control groups. PC/GPC: phosphorylcholine/glycerolphosphocholine.

Table 3 OPLS-DA coefficients derived from the NMR data of plasma metabolites from the (A) control, (B) sweet potato fiber, and (C) sweet potato residue groups

Metabolite	B (vs. A)^a	C (vs. A)^a	B (vs. C)^a
a Correlation coefficients: positive and negative signs indicate positive and negative correlation in the concentrations, respectively. The correlation coefficient |r| > 0.602 (for B vs. A and C vs. A) or 0.576 (for B vs. C) was used as the cutoff value. “—” means the correlation coefficient |r| is less than 0.602 (for B vs. A and C vs. A) or 0.576 (for B vs. C).
Acetone (15)	—	0.754	—
Citrate (19)	—	−0.698	—
Glutamate (58)	—	−0.737	—
Glutamine (61)	−0.620	−0.861	—
Glutamine/glutamate	−0.640	—	—
Isoleucine (49)	—	−0.804	—
Lactate (9)	0.658	0.792	—
Lipid (54)	0.637	0.745	—
Lysine (55)	−0.696	−0.639	−0.582
Methionine (59)	—	−0.811	—
myo-Inositol (64)	0.741	—	—
Phosphorylcholine/glycerolphosphocholine	−0.605	—	−0.701
Tyrosine (67)	−0.668	—	—
Unsaturated lipids (66)	—	0.758	—
VLDL (48)	—	0.677	—
α-Glucose (35)	−0.776	−0.680	−0.664
β-Glucose (34)	−0.778	−0.604	−0.735

---

PLS-DA was conducted on the urine spectra of the SF, SR, and control groups. The score plots (Fig. 3B) highlighted three clusters corresponding to the three groups. The metabolic profiles of the three groups were compared using OPLS-DA to further identify the key urine metabolic changes. Multivariate data analysis showed that the urine levels of acetamide, alanine, citrulline, ethanol, lactate, methylamine, methylmalonate, N-acetylglutamate, and α-hydroxybutyrate as well as the urine levels of acetate, citrulline, N-acetylglutamate, and p-hydroxyphenylacetate were higher in the SF and SR groups than in the control group (P < 0.05). By contrast, compared with the control group, the urine level of allantoin in the SF group and the urine level of isobutyrate in the SR group were lower (P < 0.05, Fig. 5A and B and Table 3). We also compared the metabolic profile of the SF group with that of the SR group using OPLS-DA. The urine levels of isobutyrate and α-hydroxybutyrate were significantly higher in the SF group than in the SR group (P < 0.05, Fig. 5C and Table 4).

Fig. 5 OPLS-DA score plots (left panel) and the corresponding coefficient loading plots (right panel) of urinary metabolites derived from the control (black squares), sweet potato fiber (green triangles), and sweet potato residue (red circles) groups (A: R²X = 25.7%, Q² = 0.529; B: R²X = 20.7%, Q² = 0.198; C: R²X = 20.8%, Q² = 0.194). One urinary sample from the sweet potato fiber group and two samples from the control group were excluded because they were positioned outside the Hotelling's T² ellipse on the score plot. The colour scale in the coefficient plot shows the significance of metabolite variation between the sweet potato fiber, sweet potato residue, and control groups.

Table 4 OPLS-DA coefficients derived from the NMR data of urine metabolites from the (A) control, (B) sweet potato fiber, and (C) sweet potato residue groups

Metabolite	B (vs. A)^a	C (vs. A)^a	B (vs. C)^a
a Correlation coefficients: positive and negative signs indicate positive and negative correlation in the concentrations, respectively. The correlation coefficient |r| > 0.666 (for B vs. A and B vs. C) or 0.632 (for C vs. A) was used as the cutoff value. “—” means the correlation coefficient |r| is less than 0.666 (for B vs. A and B vs. C) or 0.632 (for C vs. A).
Acetamide (13)	0.741	—	—
Acetate (12)	—	0.651	—
Alanine (10)	0.707	—	—
Allantoin (36)	−0.684	—	—
Citrulline (11)	0.83	0.699	—
Ethanol (6)	0.73	—	—
Isobutyrate (5)	—	−0.730	0.781
Lactate (9)	0.909	—	—
Methylamine (20)	0.724	—	—
Methylmalonate (7)	0.769	—	—
N-Acetylglutamate (14)	0.677	0.704	—
p-Hydroxyphenylacetate (39)	—	0.686	—
α-Hydroxybutyrate (3)	0.694	—	0.677
α-Hydroxy-iso-valerate (4)	—	—	0.690
α-Hydroxy-n-valerate (8)	0.75	—	—

---

Discussion

Effect of sweet potato residue and sweet potato fiber supplementation on lipid metabolism

SF supplementation can decrease lipid oxidation. Acetoacetate and 3-hydroxybutyrate are products of fatty acid oxidation in the liver, and their ratios are useful indicators of the mitochondrial redox state.²³ SF supplementation increased the urinary level of 3-hydroxybutyrate but did not change that of acetoacetate. As a result, the acetoacetate/3-hydroxybutyrate ratio decreased. This result further suggests a less oxidized cell state, which can be brought about by the decreased oxidation of fatty acids. This decrease in oxidation might be caused by a high content of antioxidants, such as ferulic acid, lignins, phytic acid, zinc, copper, selenium, and manganese in the grain envelope. Vitamin E in the germ of the fiber may have also inhibited the endogenous oxidation of lipids.²⁴ In support of this view, urinary allantoin levels were decreased by SF. Allantoin is the end product of the oxidation of uric acid by purine catabolism. Allantoin in the urine can be produced by non-enzymatic means through high levels of reactive oxygen species. Thus, allantoin can be used as a marker of oxidative stress. This finding implies that SF functions as an antioxidant. In this study, SF decreased the plasma triglyceride levels compared with the levels in the control group. Furthermore, SR can increase plasma acetone, VLDL and LDL levels while decreasing plasma triglyceride levels in rats, which may imply a change in lipid metabolism. Collectively, SF and SR can alter lipid metabolism.

Another new and intriguing observation from this study is that dietary SF supplementation changed the concentrations of lipid-signalling molecules in rat plasma. In this study, plasma glycerolphosphocholine/phosphorylcholine levels decreased and myo-inositol levels increased in response to dietary SF supplementation. myo-Inositol, a carbocyclic polyol, provides the structural basis for a number of secondary messengers (including inositol phosphates, phosphatidylinositol, and phosphatidyl inositol phosphate) in eukaryotic cells.²⁵ Therefore, inositol is related to the regulation of intracellular calcium concentrations, insulin signal transduction, gene expression, and oxidation of fatty acids.^25,26 As the main components of biological membranes, inositol and glycerolphosphocholine/phosphorylcholine concentrations are associated with the structural components of cell membranes.^27,28 We infer that the changes in these compounds could represent altered membrane turnover.

Effect of sweet potato residue and sweet potato fiber supplementation on glucose and energy metabolism

SF and SR exposure can promote glycolysis and decrease glycogenolysis. In this study, SF and SR significantly decreased plasma glucose compared with the control group. Glucose is a major substrate that offers energy for animal growth and development. Moreover, SF can increase urinary alanine levels. Alanine is an important participant and regulator in glucose metabolism. Taken together, SF can alter the glucose–alanine cycle metabolism. Increased lactate concentration was also observed in the urine and plasma of SF-supplemented groups and in the plasma of the SR-supplemented groups. Lactate is the end product of compounds related to energy metabolism. Increased lactate level is involved in increased anaerobic glycolysis. In addition, an increased plasma lactate level implies inhibited gluconeogenesis as well as changed carbohydrate and energy metabolism. SR can also decrease urinary citrate levels compared with the control group, implying that the tricarboxylic acid cycle was altered by SR in rats. Collectively, SF and SR can alter energy metabolism.

Effect of sweet potato residue and sweet potato fiber supplementation on amino acid metabolism

SF and SR exposure can alter the urea cycle. In this study, urinary citrulline and N-acetylglutamate levels were increased. Citrulline is an amino acid made from ornithine and carbamoyl phosphate in one of the central reactions in the urea cycle. Citrulline is also obtained from arginine as a by-product of the reaction catalyzed by the nitric oxide synthases family. In this reaction, arginine is first oxidized into N-hydroxyl-arginine, which is further oxidized to citrulline, accompanied by the release of nitric oxide. Urea has a significant role in the metabolism of nitrogen-containing compounds. N-Acetylglutamate is essential for the normal function of the urea cycle, and the variations in its concentration influence the urea production rate along with the other substrates for urea synthesis.²⁹ Compared with the control group, the SF group exhibited decreased plasma blood urea nitrogen levels. The decrease in blood urea nitrogen is linked with the increase in N-acetylglutamate. In this study, plasma tyrosine and total protein, which are both involved in protein synthesis, were decreased by SF. More amino acids were also decreased in the protein synthesis, resulting in lower levels of amino acids in the plasma. In this study, we observed a reduction in the levels of plasma lysine, glutamate, and glutamine/glutamate, in accordance with the role of SF in decreasing protein synthesis in the rats. Plasma glutamate, glutamine, lysine, and methionine levels decreased in the SR group, compared with the control group. Changes in the levels of plasma metabolites (glutamine, lysine, and methionine) that could be involved in the modulation of immunological responses were also observed in the presence of the probiotic treatment. Glutamine can activate signalling pathways to enhance protein synthesis, supporting animal growth and development. Collectively, these observations imply that SF supplementation can inhibit protein synthesis. Our results also demonstrated that the levels of branched-chain amino acids (isoleucine) were decreased by SR supplementation. The possible reason for this is that increases in energy expenditure can also cause the elevated consumption of amino acids such as isoleucine to provide energy. Therefore, SF and SR supplementation can change amino acid metabolism.

Effect of sweet potato residue and sweet potato fiber supplementation on gut microbiota metabolism

SF and SR exposure can alter the gut microbiota metabolism. Urinary concentrations of ethanol and nitrogenous product (methylamine) were decreased in the SF group. Isobutyrates and acetate, both short-chain fatty acids, were decreased and increased, respectively, in the SR group. Notably, these compounds are microbial metabolites of carbohydrates and amino acids, which are likely produced in the lumen of the small and large intestines.^30,31 The altered levels of gut microbial cometabolites including p-hydroxyphenylacetate were also associated with the disturbance in gut microbiota due to SR exposure. Phenylacetate was transformed from phenylalanine by the action of gut microbiota, and phenylacetate was then conjugated with glycine to form phenylacetylglycine.³² Previous reports have suggested that elevated levels of urinary phenylacetylglycine with abnormal accumulations of phospholipids in rat liver, and that these levels can serve as a surrogate biomarker for associated changes in gut microbiota. Acyl-CoA has a major function in glycine conjugation,³³ but whether or not this enzyme is regulated by SR exposure remains unclear. Moreover, p-hydroxyphenylacetate is a metabolite of tyrosine through enteric bacteria, and mammalian metabolism is significantly affected by its interaction with the complex gut microbial community. This finding may indicate that certain compounds in urine such as p-hydroxyphenylacetate are unique metabolic products of the microbes that inhabit the lumen of the gut, which is strictly dependent on diet. Changes in these metabolites can be attributed to the reduced number and/or altered activity of intestinal microorganisms. Gut microbiota significantly affect the development and structure of the intestinal epithelium, the digestive and absorptive properties of the intestine, and the host immune system.³⁴ Possible disturbances of gut microbiota by SF and SR administration can affect health; thus, microbiological identification of specific changes in the microbiota community can help address the metabolic implications of SR and SF supplementation.

Conclusion

SF and SR supplementation affects the urine and plasma metabolome of rats, and can induce common systemic metabolic changes in lipid metabolism, glycogenolysis and glycolysis metabolism, energy metabolism, protein biosynthesis, and gut microbiota metabolism. Our study has shown that the metabolomic strategy is useful in examining the effects of nutritional intervention in a mammalian system. To the best of our knowledge, this study is the first in vivo report on the response of animal biological systems to SF and SR supplementation. Future studies may focus on a mechanistic understanding of SF and SR effects on animal tissue intermediary metabolism.

Acknowledgements

We would like to thank the staff at our laboratory for their on-going assistance. This investigation received financial support from the National Natural Science Foundation of China (no.: 31301986), the Research Foundation of Education Bureau of Sichuan Province (no.: 00924901), the Ministry of Education Chunhui Project of China (no.: Z2010092) and the Specific Research Supporting Program for Discipline Construction in Sichuan Agricultural University.

References

1. American Association of Cereal Chemists, Cereal Foods World, 2001, vol. 46, pp. 112–126.
2. T. Happi Emaga, C. Robert, S. N. Ronkart, B. Wathelet and M. Paquot, Bioresour. Technol., 2008, 99, 4346–4354.
3. A. Nawirska and M. Kwasniewska, Food Chem., 2005, 91, 221–225.
4. N. Vergara-Valencia, E. Granados-Pérez, E. Agama-Acevedo, J. Tovar, J. Ruales and L. A. Bello-Pérez, LWT–Food Sci. Technol., 2007, 40, 722–729.
5. A. M. Pylkas, L. R. Juneja and J. L. Slavin, J. Med. Food, 2005, 8, 113–116.
6. J. Chen, T. Gislettle and X. Huang, Am. J. Clin. Nutr., 2010, 92, 1538.
7. S. M. Gråsten, A. Pajari, K. Liukkonen, S. Karppinen and H. M. Mykkänen, Nutr. Res., 2002, 22, 1435–1444.
8. A. Redondo-Cuenca, M. J. Villanueva-Suárez, M. D. Rodríguez-Sevilla and I. Mateos-Aparicio, Food Chem., 2007, 101, 1216–1222.
9. D. Betancur-Ancona, G. Peraza-Mercado, Y. Moguel-Ordeňez and S. Fuertes-Blanco, Food Chem., 2004, 84, 287–295.
10. X. Mei, T. H. Mu and J. J. Han, J. Agric. Food Chem., 2010, 58, 7305–7310.
11. T. P. Trinidad, R. S. Sagum, A. C. Mallillin, M. S. Borlagdan, M. P. de Leon and T. F. Aviles, Food Nutr. Sci., 2013, 4, 491–495.
12. J. K. Nicholson and I. D. Wilson, Nat. Rev. Drug Discovery, 2003, 2, 668–676.
13. C. Stella, B. Beckwith-Hall, O. Cloarec, E. Holmes, J. C. Lindon, J. Powell, F. van de Ouderaa, S. Bingham, A. J. Cross and J. K. Nicholson, J. Proteome Res., 2006, 5, 2780–2788.
14. J. K. Nicholson, E. Holmes, J. C. Lindon and I. D. Wilson, Nat. Biotechnol., 2004, 22, 1268–1274.
15. G. M. Liu, T. Yan, J. Wang, Z. Q. Huang, X. L. Chen, G. Jia, C. M. Wu, H. Zhao, B. Xue, L. Xiao and J. Y. Tang, J. Agric. Food Chem., 2013, 61, 11212–11221.
16. P. G. Reeves, F. H. Nielsen and G. C. Fahey Jr, J. Nutr., 1993, 123, 1939–1951.
17. M. M. Cassidy, F. G. Lightfoot, L. E. Grau, J. A. Story, D. Kritchevsky and G. V. Vahouny, Am. J. Clin. Nutr., 1981, 34, 218–228.
18. J. K. Nicholson and P. J. D. Foxall, Anal. Chem., 1995, 67, 793–811.
19. T. W. M. Fan, Prog. Nucl. Magn. Reson. Spectrosc., 1996, 28, 161–219.
20. J. Trygg and S. Wold, J. Chemom., 2002, 16, 119–128.
21. O. Cloarec, M. E. Dumas, J. Trygg, A. Craig, R. H. Barton, J. C. Lindon, J. K. Nicholson and E. Holmes, Anal. Chem., 2005, 77, 517–526.
22. F. Lindgren, B. Hansen, W. Karcher, M. SjÖstrÖm and L. Eriksson, J. Chemom., 1996, 10, 521–532.
23. G. Y. Wu, A. Gunasekara, H. Brunengraber and E. B. Marliss, Am. J. Physiol.: Endocrinol. Metab., 1991, 261, E211–E226.
24. J. L. Slavin, M. C. Martini, D. R. Jacobs Jr and L. Marquart, Am. J. Clin. Nutr., 1999, 70, S459–S463.
25. G. Di Paolo and P. De Camilli, Nature, 2006, 443, 651–657.
26. R. A. Siddiqui, K. A. Harvey and G. P. Zaloga, J. Nutr. Biochem., 2008, 19, 417–437.
27. J. Klein, J. Neural Transm., 2000, 107, 1027–1063.
28. S. H. Zeisel and J. K. Blusztajn, Annu. Rev. Nutr., 1994, 14, 269–296.
29. A. J. Meijer and A. J. Verhoeven, Biochem. J., 1984, 223, 559–560.
30. S. Rezzi, Z. Ramadan, L. B. Fay and S. Kochhar, J. Proteome Res., 2007, 6, 513–525.
31. G. W. Tannock, Appl. Environ. Microbiol., 2004, 70, 3189–3194.
32. E. Bohus, M. Coen, H. C. Keun, T. M. Ebbels, O. Beckonert, J. C. Lindon, E. Holmes, B. Noszál and J. K. Nicholson, J. Proteome Res., 2008, 7, 4435–4445.
33. F. Kasuya, K. Igarashi and M. Fukui, Biochem. Pharmacol., 1996, 51, 805–809.
34. J. Xu and J. I. Gordon, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 10452–10459.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra02421d

‡ Contributed equally.

---