Investigation of Andrographolide Effect on Non-Infected Red Blood Cells Using the 1H-NMR-Based Metabolomics Approach

H OH metabolites OH

Article Investigation of Andrographolide Effect on Non-Infected Red Blood Cells Using the 1H-NMR-Based Metabolomics Approach

Ashraf Ahmad Issa Alapid 1,2 , Roslaini Abd. Majid 3, Zaid O. Ibraheem 4, Ahmed Mediani 5 , Intan Saﬁnar Ismail 6 , Ngah Zasmy Unyah 1, Sharif Alhassan Abdullahi 7, Norshariza Nordin 8 , Mohammed Nasiru Wana 9 and Rusliza Basir 10,*

1 Medical Parasitology Unit, Department of Medical Microbiology and Parasitology, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, Serdang 43400, Selangor, Malaysia; [email protected] (A.A.I.A.); [email protected] (N.Z.U.) 2 Department of Zoology, Faculty of Science-Alasaba, University of Gharyan, Gharyan 010101, Libya 3 Faculty of Medicine and Defence Health, National Defence University of Malaysia, Kuala Lumpur 57000, Malaysia; [email protected] 4 Department of Pharmacy, Al Rafidain University College, Al Mustansyria, Baghdad 10052, Iraq; [email protected] 5 Institute of Systems Biology (INBIOSIS), Universiti Kebangsaan Malaysia, Bangi 43600, Selangor, Malaysia; [email protected] 6 Natural Medicine and Products Research Laboratory, Institute of Bioscience, Universiti Putra Malaysia, Serdang 43400, Selangor, Malaysia; safi[email protected] 7 Department of Medical Microbiology and Parasitology, Faculty of Clinical Sciences, Bayero University Kano, Citation: Alapid, A.A.I.; Abd. Majid, Kano 700241, Nigeria; [email protected] R.; Ibraheem, Z.O.; Mediani, A.; 8 Department of Biomedical Sciences, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, Ismail, I.S.; Unyah, N.Z.; Alhassan Serdang 43400, Selangor, Malaysia; [email protected] 9 Abdullahi, S.; Nordin, N.; Nasiru Department of Biological Sciences, Faculty of Science, Abubakar Tafawa Balewa University Bauchi, Wana, M.; Basir, R. Investigation of Bauchi 740272, Nigeria; [email protected] 10 Andrographolide Effect on Pharmacology Unit, Department of Human Anatomy, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, Serdang 43400, Selangor, Malaysia Non-Infected Red Blood Cells Using * Correspondence: [email protected]; Tel.: +60-124-747-459 the 1H-NMR-Based Metabolomics Approach. Metabolites 2021, 11, 486. Abstract: https://doi.org/10.3390/metabo Andrographolide (AG) has been shown to have several medicinal and pharmaceutical 11080486 effects, such as antimicrobial, anti-inflammatory, antioxidant, anti-diabetic, and anti-malarial activities. Moreover, studies to assess the pharmacological effect of AG on the metabolic changes of Academic Editor: uninfected red blood cells (uRBCs) have not yet been investigated. This study aims to evaluate Anthony Tsarbopoulos the pharmacological effects of AG compared to chloroquine (CQ) on the metabolic variations of uRBCs in vitro using a proton nuclear magnetic resonance (1H-NMR)-based metabolomics approach Received: 8 June 2021 coupled with multivariate data analysis (MVDA). Forty-one metabolites were successfully iden- Accepted: 15 July 2021 tified by 1H-NMR. The results of the unsupervised data analysis principal component analysis Published: 28 July 2021 (PCA) showed ideal differentiation between AG and CQ. PC1 and PC2 accounted for 71.4% and 17.7% of the explained variation, respectively, with a total variance of 89.10%. Based on S-plot Publisher’s Note: MDPI stays neutral and VIP values, a total of 28 and 32 metabolites were identified as biomarkers in uRBCs-AG and with regard to jurisdictional claims in uRBCs-CQ, respectively. In uRBCs treated with AG, ten metabolic pathways were determined to be published maps and institutional affil- disturbed, including riboflavin metabolism, D-glutamate and D-glutamine metabolism, phenylala- iations. nine metabolism, glutathione metabolism, proline and arginine metabolism, arginine biosynthesis, citrate cycle, glycolysis/gluconeogenesis, and pyruvate metabolism as well as alanine, aspartate, and glutamate metabolism. In contrast, in CQ-treated uRBCs, nine affected metabolic pathways were determined, which involved the same metabolic pathways for uRBCs-AG, except for glutathione Copyright: © 2021 by the authors. metabolism. These findings suggest an evident relationship between AG and CQ associated with Licensee MDPI, Basel, Switzerland. This article is an open access article metabolic changes in intact RBCs after being exposed to the treatment. The metabolomics results distributed under the terms and could allow useful comprehensive insights into the underlying mechanism of the action of AG and 1 conditions of the Creative Commons CQ on red blood cells. Consequently, the H-NMR-based metabolomics approach was successfully Attribution (CC BY) license (https:// utilized to identify the pharmacological effects of AG and CQ on the metabolic variations of uRBCs. creativecommons.org/licenses/by/ 4.0/).

Metabolites 2021, 11, 486. https://doi.org/10.3390/metabo11080486 https://www.mdpi.com/journal/metabolites Metabolites 2021, 11, 486 2 of 24

Keywords: andrographolide; chloroquine; uninfected RBCs; 1H-NMR; metabolic pathway; anti-malarial

1. Introduction Human red blood cells (RBCs), also known as erythrocytes, represent a high number of types of blood cells in the human body. They also account for nearly half of the total blood volume and their diameters are 7–8 µm [1]. Normal human RBCs are composed of high concentrations of different metabolites that act as coenzymes of redox reactions, such as nicotinamide adenine dinucleotide phosphate (NADP+ and NADPH) and oxi- dized/reduced nicotinamide adenine dinucleotide (NAD+ and NADH). In addition, they contain enzymatic antioxidants, such as catalase (Cat), superoxide dismutase (SOD), glutathione peroxidase (GPx), and peroxiredoxin 2 (Prx2). The coenzymes of energy are also present, such as adenosine diphosphate (ADP), adenosine monophosphate (AMP), and adenosine triphosphate (ATP), as well as non-enzymatic antioxidants, such as oxi- dized and reduced glutathione (GSSG and GSH). All of these components can be found in plasma/serum [2,3]. Unlike several other cells in the human body, mature RBCs lack nuclei, mitochondria, ribosomes, and other organelles, which are unable to generate energy through the Krebs cycle, and synthesize new proteins [4,5]. For this reason, the anaerobic degradation of glucose via the glycolytic pathway known as the Embden–Meyerhof pathway (EMP) is the only source of produced ATP and NADH [6]. It is very important to note that RBCs are highly metabolically active cells and glycolysis is the main metabolic pathway in RBCs that provides the energy demand. They also have a function to protect cells against reactive oxygen species (ROS) using the GSH antioxidant [2]. RBCs, as the most abundant types of blood cells, work as a perfect candidate for in vivo drug delivery due to many characteristics, including an extended circulation period of ~120 days, flexibility, superb biocompatibility, and minimal immunogenicity [7–10]. Due to these remarkable properties, RBCs have been widely studied as drug delivery systems [11]. Metabolomics can study the profiling of the biochemical compounds or metabolites which are changed by the physiological or pathological state in cells, tissue, biofluids, organs, and organisms [12,13]. Many analytical tools have been employed in metabolomics studies, of which proton nuclear magnetic resonance (1H-NMR) spectroscopy is the major one. Its significant role is in the analysis of the metabolites in different biological samples, including RBCs, serum, plasma, and urine [14]. Yet, blood analysis using metabolomics is limited to the analysis of either serum or plasma metabolites only, while all other blood components obtained during sample preparation, including the RBCs, white blood cells (WBCs), and platelets (PLT) are barely examined [2]. So far, only a handful of researchers have studied the metabolomics variation of RBCs and infected RBCs with different species of Plasmod- ium parasites using 1H-NMR coupled with multivariate data analysis (MVDA) [2,15–17]. In addition, metabolomics can be used to evaluate the changes in metabolites of blood cells upon treatment with herbal drugs. Andrographis paniculate is one of the medical plants which has been used extensively in Southern Asia, China, and India for many clinical applications [18]. A. paniculate has many bioactive compounds, one of which is andrographolide (AG). It is a labdane diterpenoid derivative reported to exhibit many medicinal and pharmaceutical attributes, such as antimicrobial, anti-inflammation, antioxidant [19], cardio-protection, hepatoprotection [20], anti-HIV, anti-carcinogenic [21,22], anti-diabetic, anti-trypanosomal activity [23], and antimalarial [24–31]. Concerning the effect of AG as an anti-malarial drug, there is one study that has reported the impact of AG on the membranes of uninfected RBCs (uRBCs) and infected RBCs (iRBCs) by Plasmodium falciparum 3D7. The results showed that AG was harmless to uRBCs at concentrations approach in its medicinal effect on Plasmodia. How- ever, this stability declined at higher concentrations [30]. In iRBCs, a high concentration of AG inhibits the Plasmodium-induced permeation pathway as well as the capability of Metabolites 2021, 11, 486 3 of 24

merozoites to infect new RBCs [30]. In this context, AG has many pharmacological activities, when used as an anti-malarial drug on both uRBCs and iRBCs with the Plasmodium parasite. More significantly, mature RBCs do not have any genetic elements, making them less risky compared to other cell therapies and genes [5]. Undoubtedly, AG may have an effect on the metabolism of uRBCs and iRBCs. However, there is no study of the AG effects on the metabolism of uRBCs of the host, especially when using it as an anti-malarial drug. Nevertheless, many studies suggested that RBCs might contribute to the metabolism of exogenous and endogenous elements, including drugs [32,33]. Consequently, understand- ing the impact of these elements or drugs on intact RBC metabolism and its biology may have critical outcomes in the field of antimalarial drug discovery. Metabolites of RBC as the downstream of gene expression can show the real changes upon treatment with drugs. Metabolites have numerous activities in cell physiology, including fuel, structure, signaling, stimulating, and inhibiting enzyme effects; catalytic activity as an enzyme cofactor, defense, and interactions with others. To the best of our knowledge, this study was carried out to fill up this gap. In addition, there has been no study conducted to investigate and interpret the metabolic changes associated with AG on uRBCs. Therefore, this study aimed to evaluate the metabolic variations of uRBCs following exposure to AG and compared with chloroquine (CQ) by using a 1H-NMR-based metabolomics approach. It is essential to assess the effects of AG on uRBCs in vitro when intending to reveal some potent insights into the field of antimalarial drugs with a novel mechanism of action.

2. Results and Discussion 2.1. 1H-NMR Spectra of Uninfected RBC Extracts The representative 500 MHz 1H-NMR spectra of the extracts of uRBCs-untreated, uRBCs-AG and uRBCs-CQ are shown in Figure1. By examining the spectra of the three samples, there were variations in the intensity of the peak signals, wherein the uRBCs- AG revealed a lower intensity compared to the those of uRBCs-CQ as well as uRBCs- control (Figure1). The metabolite signals were then assigned based on the previous studies [2,34], Human Metabolome Database (HMDB), and the library of Chenomx NMR suite 7.5 (Chenomx Inc., Canada) by comparison with the 1H-NMR signals of reference compounds. Human red blood cells contain a wealth of metabolites, such as amino acids, carbohydrates, fatty acids, and coenzymes. In addition, RBCs represent a high percentage of total blood cells [2,34]. In this study, 41 blood metabolites were identified as shown in Table1 . All these metabolites have been previously reported in many studies [16,17,35,36]. The metabolites belonged to different chemical classes, such as fatty acids, amino acids, organic acids, and carbohydrates. The first region of the spectra from 0.9 to 3.5 ppm exhibited signals that primarily belong to fatty acids (i.e., isovalerate and 3-hydroxybutyric) as well as amino acids (i.e., valine, lysine, alanine, arginine, threonine, and glutamine). The middle region of the spectra (3.5–5.6 ppm) showed the signals of carbohydrates (i.e., fructose, glucose, and ribose) and some amino acid signals (i.e., glycine, threonine, arginine. The remaining part of the spectra (5.7–9.0 ppm) showcased signals of organic acids and purine nucleotides (i.e., lactate, isocitric acid, adenosine monophosphate (AMP), and adenosine triphosphate (ATP). Table1 summarizes all the identified metabolites of uninfected RBC extracts, and their characteristic signals. Metabolites 2021, 11, x FOR PEER REVIEW 4 of 26 Metabolites 2021, 11, 486 4 of 24

FigureFigure 1. 1.1H-NMR 1H-NMR spectra spectra (500(500 MHz, CPMG) CPMG) of of untreated untreated uRBCs uRBCs-control-control (A), ( AuRBCs), uRBCs-AG-AG (B), and (B), uRBCs and uRBCs-CQ-CQ (C). The (C 1).H The- 1H-NMRNMR signals signals identified identified were: were: 1; 1;isovalerate isovalerate,, 2; 2 2;-aminobutyric 2-aminobutyric acid acid,, 3; valine 3; valine,, 4; isobutyrate 4; isobutyrate,, 5; 3-hydroxybutyrate 5; 3-hydroxybutyrate,, 6; la- 6; la-tate,tate, 7;7; alanine,alanine, 8 8;; lysine lysine,, 9 9;; ornithine ornithine,, 10 10;; proline proline,, 11 11;; homoserine homoserine,, 12 12;; methionine methionine,, 13 13;; acetone acetone,, 14; 14; glutamate glutamate,, 15; 15;pyruvate pyruvate,, 16; l-glutamine, 17; riboflavin, 18; citrate, 19; glutathione, 20; creatine, 21; cis-aconitic acid, 22; choline, 23; glucose, 24; 16; l-glutamine, 17; riboflavin, 18; citrate, 19; glutathione, 20; creatine, 21; cis-aconitic acid, 22; choline, 23; glucose, 24; glycine, 25; threonine, 26; arginine, 27; betaine, 28; ribose, 29; mannose, 30; fructose, 31; isocitrate, 32; pyroglutamic acid, glycine,33; n- 25;acetylcysteine threonine,, 26;34; arginine,2-phosphoglycerate 27; betaine,, 35 28;; guanosine ribose, 29; triphosphate mannose, 30; ( fructose,GTP), 36; 31;4-pyridoxate isocitrate,, 32;37; pyroglutamicadenosine triphos- acid, 33; n-acetylcysteine,phate (ATP), 38 34;; histidine 2-phosphoglycerate,, 39; 3-methylhistidine 35; guanosine, 40; adenosine triphosphate monophosphate (GTP), 36; (AMP 4-pyridoxate,), 41; phenylalanine. 37; adenosine triphosphate (ATP), 38; histidine, 39; 3-methylhistidine, 40; adenosine monophosphate (AMP), 41; phenylalanine. Table 1. 1H-NMR characteristic signals of the identified metabolites in the red blood cells. Table 1. 1 No. HMDB ID H-NMRMetabolites characteristic signals of the1H- identifiedNMR Signals metabolites in the red blood cells. 1. HMDB0000718 No. HMDBIsovalerate ID Metabolites 0.94 (d, 6.6),1H-NMR 1.93 (m) Signals 1. HMDB0000718 Isovalerate 0.94 (d, 6.6), 1.93 (m) 2. HMDB0000452 2-aminobutyric acid 0.98 (t, 7.6) 2. HMDB0000452 2-aminobutyric acid 0.98 (t, 7.6) 3. HMDB0000883 3. HMDB0000883Valine Valine1.02 (d, 7.1), 1.02 3.60 (d, (d, 7.1), 4.4) 3.60 (d, 4.4) 4. HMDB0001873 Isobutyrate 1.06 (d, 6.8) 4. HMDB0001873 5. HMDB0000011Isobutyrate 3-Hydroxybutyrate1.06 (d, 6.8) 1.20 (d, 6.2) 6. HMDB0000190 Lactate 1.34 (d, 7.0) 5. HMDB0000011 7. HMDB00001613-Hydroxybutyrate Alanine 1.20 (d, 6.2) 1.48 (d, 7.3) 8. HMDB0000182 Lysine 1.74 (m) 6. HMDB0000190 9. HMDB0000214Lactate Ornithine1.34 (d, 7.0) 3.06 (t, 7.5), 1.78 (m) 7. 10. HMDB0000162 Proline 1.99 (m), 3.33 (m), 2.06 (m), 4.13 (dd, 6.4, 8.7) HMDB0000161 11. HMDB0000719Alanine Homoserine1.48 (d, 7.3) 2.02 (m) 8. 12. HMDB0000696 Methionine 2.14 (m) HMDB0000182 13. HMDB0001659Lysine Acetone1.74 (m) 2.22 (s) 14. HMDB0000148 Glutamate 2.38 (m), 2.32 (m) 9. HMDB0000214 Ornithine 3.06 (t, 7.5), 1.78 (m) 15. HMDB0000243 Pyruvate 2.36 (s) 10. HMDB0000162 16. HMDB0000641Proline L-Glutamine1.99 (m), 3.33 2.44 (m), (m) 2.06 (m), 4.13 (dd, 6.4, 8.7) 17. HMDB0000244 Riboflavin 2.49 (s), 2.56 (s) 11. HMDB0000719 18. HMDB0000094Homoserine Citrate2.02 (m) 2.66 (d, 15.1) 19. HMDB0000125 Glutathione 2.18 (m), 2.58 (m), 2.98 (m) 12. HMDB0000696 20. HMDB0000064Methionine Creatine2.14 (m) 3.02 (s) 21. HMDB0000072 Cis-Aconitic acid 3.10 (s)

13. HMDB0001659 22. HMDB0000097Acetone Choline2.22 (s) 3.22 (s) 23. HMDB0003345 Glucose 5.22 (d, 3.8), 3.40 (m), 3.45 (m) 14. HMDB0000148 24. HMDB0000123Glutamate Glycine2.38 (m), 2.32 3.54 (m) (s) 15. 25. HMDB0000167 Threonine 3.58 (d, 4.9) HMDB0000243 26. HMDB0000517Pyruvate Arginine2.36 (s) 3.78 (t, 6.5) 27. HMDB0000043 Betaine 3.89 (s) 28. HMDB0000283 Ribose 3.99 (m), 3.82 (m)

Metabolites 2021, 11, 486 5 of 24

Table 1. Cont.

No. HMDB ID Metabolites 1H-NMR Signals 29. HMDB0000169 Mannose 5.18 (d, 1.7), 3.86 (m) 30. HMDB0000660 Fructose 4.01 (m) 31. HMDB0000193 Isocitrate 2.95 (m) 32. HMDB0000267 Pyroglutamic acid 2.50 (m), 4.16 (dd, 7.5, 5.0) 33. HMDB0001890 N-Acetylcysteine 4.36 (m) 34. HMDB0003391 2-Phosphoglycerate 4.50 (m) 35. HMDB0001273 GTP 4.55 (d,5.0) 36. HMDB0000017 4-Pyridoxate 4.74 (s) 37. HMDB0000538 ATP 6.13 (d, 5.7) 38. HMDB0000177 Histidine 7.06 (s) 39. HMDB0000479 3-Methylhistidine 7.92 (s) 40. HMDB0000045 AMP 8.23 (s), 8.58 (s) 41. HMDB0000159 Phenylalanine 3.19 (m), 7.32 (d, 7.4) The small letters in parentheses signify: s; singlet, d; doublet, dd; doublet of doublet, t; triplet, and m; multiplet. J couplings in Hz. HMDB; Human Metabolome Database, ATP; adenosine triphosphate, AMP; adenosine monophosphate, GTP; guanosine triphosphate.

2.2. Principal Component Analysis (PCA) of 1H-NMR Data Principal component analysis (PCA) is one of the most practical and reliable MVDA methods. As an unsupervised method, PCA is employed to minimize the dimensionality of the multivariate dataset. PCA is generally presented graphically as a score and loading plots that are used to determine the similarities or the metabolic differences between biological samples. Therefore, one of the benefits of PCA is that it can help in showing the occurrence of clusters or outliers between the samples represented in the score plot. In order to determine the metabolites that have contributed to a separation in the score plot, the loading plot is used. This plot shows the variables that are responsible for the discernment pattern between the samples as well as the recognition of spectral signals resulting from the discovery of the compounds [37]. The binned 1H-NMR data were subjected to PCA in order to distinguish the variations or similarities of the blood metabolites among the treated and untreated uninfected RBCs samples. The PCA model showed a clear separation between three groups into three clusters (Figure2A). The model has great goodness of fit, with R2Y = 0.968, and good predictability, with Q2 = 0.906. In the PCA score plot, the first two principal components, PC1 and PC2, revealed a total variance of 89.10%, with PC1 accounting for 71.4% and PC2 for 17.4%. The control group was separately discriminated from the other two treated groups with AG and CQ by PC1. The uRBCsCQ group was separated from the control and uRBCsAG by PC2 (Figure2A). No outlier data were observed in the X-space based on the DModX analysis [38]. The loading scatter plot was performed in order to know the characteristic variables that contributed to the separation in score plot. Consequently, both the loading and score plots are complementary to each other and aid the comprehensive insight into the differentiation between the untreated uRBCs, uRBCs-AG and uRBCs-CQ (Figure2B). The uRBCs were located on the negative side of PC1, characterized as having lower level/concentration of most blood metabolites, while the two uRBCs treated groups (uRBCs-AG and uRBCs-CQ) were on the positive side of PC1 with higher levels of blood metabolites. The metabolite signals in the upper-right quadrant of the scatter plot contribute to the separation of uRBCs- AG, including ribose, arginine, glucose, acetone, citrate, 4-pyridoxate, 3-hydroxybutyrate, and riboflavin. On the other hand, in the uRBCs-CQ group, the contributing metabolite signals were in the lower-right quadrant of the scatter plot belonging to the metabolites, such as 2-aminobutyric acid, valine, isobutyrate, lactate, alanine, lysine, glutathione, and glutamate. The results of the loading scatter plot indicate that the metabolites of the uRBCs-CQ group were higher compared to uRBCs-AG (Figure2B). The observed higher level of metabolites in both treated groups could be attributed to AG and CQ. Both AG and CQ showed significant effects by shifting away from the untreated group. A possible explanation of these effects or changes in the normal components of RBCs might be the Metabolites 2021, 11, x FOR PEER REVIEW 6 of 26

occurrence of clusters or outliers between the samples represented in the score plot. In order to determine the metabolites that have contributed to a separation in the score plot, the loading plot is used. This plot shows the variables that are responsible for the discernment pattern between the samples as well as the recognition of spectral signals resulting from the discovery of the compounds [37]. The binned 1H-NMR data were subjected to Metabolites 2021, 11, 486 PCA in order to distinguish the variations or similarities of the blood metabolites among6 of 24 the treated and untreated uninfected RBCs samples. The PCA model showed a clear separation between three groups into three clusters (Figure 2A). The model has great goodness of fit, with R2Y = 0.968, and good predictability, with Q2 = 0.906. In the PCA score plot,immediate the first result two principal of an established components, pharmacologic PC1 and PC2, action revealed of this druga total [39 variance]. In addition, of 89.10%, AG withwas possiblyPC1 accounting innocent for to 71.4% RBCs atand concentrations PC2 for 17.4%. of itsThe therapeutic control group level was against separately plasmodia, dis- criminatedwhich can befrom reduced the other at higher two treated concentrations groups with [30]. AG To ﬁndand theCQ biomarkersby PC1. The for uRBCsCQ both AG groupand CQ, was separate separated models from forthe each control group and of uRBCsAG uRBCs-AG by and PC2 uRBCs-CQ (Figure 2A). were No constructedoutlier data wereby comparing observed themin the with X-space the uRBCsbased on control the DModX group. analysis [38].

FigureFigure 2. Principal component component analysis analysis ( (PCA)PCA) of of the the uninfected uninfected red red blood blood cells cells’’ metabolic profiles proﬁles:: ((A)) s scorecore scatter plot (PC1 vs. PC2) PC2),, ( (B)) the the loading loading scatter scatter plot plot,, uRBCs uRBCs;; u uninfectedninfected red red blood cells cells,, AGAG;; andrographolide andrographolide,, CQ CQ;; chloroquine.

2.3. BiomarkerThe loading Identification scatter plot by was OPLS-DA performed in order to know the characteristic variables that contributedIn most of the to the metabolomics separation in studies score conductedplot. Consequently to determine, both biomarkers, the loading OPLS-DAand score is the preferred method and has proven to be a very helpful tool for this purpose. In the OPLS-DA score plot, S-plot and variable importance in projection (VIP) values help in assigning the significant metabolites, which contribute to the biological outcomes [40]. Generally, variables with VIP values equal to or higher than 1.0 are considered to be significant features in the OPLS-DA model [38]. In this study, the OPLS-DA was employed on blood 1H-NMR data to reveal distinct observations about the significant metabolites in the models of uRBCs-AG versus uRBCs and uRBCs-CQ versus uRBCs. The score plot of OPLS-DA revealed an obvious separation by PC1 between uRBCs and uRBCs-AG. The Metabolites 2021, 11, x FOR PEER REVIEW 7 of 26

plots are complementary to each other and aid the comprehensive insight into the differentiation between the untreated uRBCs, uRBCs-AG and uRBCs-CQ (Figure 2B). The uR- BCs were located on the negative side of PC1, characterized as having lower level/concentration of most blood metabolites, while the two uRBCs treated groups (uRBCs-AG and uRBCs-CQ) were on the positive side of PC1 with higher levels of blood metabolites. The metabolite signals in the upper-right quadrant of the scatter plot contribute to the separation of uRBCs-AG, including ribose, arginine, glucose, acetone, citrate, 4-pyridoxate, 3- hydroxybutyrate, and riboflavin. On the other hand, in the uRBCs-CQ group, the contributing metabolite signals were in the lower-right quadrant of the scatter plot belonging to the metabolites, such as 2-aminobutyric acid, valine, isobutyrate, lactate, alanine, lysine, glutathione, and glutamate. The results of the loading scatter plot indicate that the metabolites of the uRBCs-CQ group were higher compared to uRBCs-AG (Figure 2B). The observed higher level of metabolites in both treated groups could be attributed to AG and CQ. Both AG and CQ showed significant effects by shifting away from the untreated group. A possible explanation of these effects or changes in the normal components of RBCs might be the immediate result of an established pharmacologic action of this drug [39]. In addition, AG was possibly innocent to RBCs at concentrations of its therapeutic level against plasmodia, which can be reduced at higher concentrations [30]. To find the biomarkers for both AG and CQ, separate models for each group of uRBCs-AG and uR- BCs-CQ were constructed by comparing them with the uRBCs control group.

2.3. Biomarker Identification by OPLS-DA In most of the metabolomics studies conducted to determine biomarkers, OPLS-DA is the preferred method and has proven to be a very helpful tool for this purpose. In the OPLS-DA score plot, S-plot and variable importance in projection (VIP) values help in assigning the significant metabolites, which contribute to the biological outcomes [40]. Generally, variables with VIP values equal to or higher than 1.0 are considered to be sig- Metabolites 2021, 11, 486 nificant features in the OPLS-DA model [38]. In this study, the OPLS-DA was employed 7 of 24 on blood 1H-NMR data to reveal distinct observations about the significant metabolites in the models of uRBCs-AG versus uRBCs and uRBCs-CQ versus uRBCs. The score plot of OPLS-DA revealed an obvious separation by PC1 between uRBCs and uRBCs-AG. The uRBCs group was on the positive side of PC1, while the group treated with AG was on the uRBCs group was on the positive side of PC1, while the group treated with AG was on negative side (Figure3A). the negative side (Figure 3A).

Metabolites 2021, 11, x FOR PEER REVIEW 8 of 26

FigureFigure 3. 3.TheThe OPLS OPLS-DA-DA score score scatter scatter plot plot(A) and (A) S and-plot S-plot (B) of (theB) ofblood the metabolic blood metabolic profiles proﬁles of unin- of fecteduninfected RBC samples. RBC samples. uRBCs uRBCs;; uninfected uninfected red blood red cells blood, AG cells,; andrographolide. AG; andrographolide.

TheThe S- S-plotplot is is the the covariance covariance and and association association loading loading diagnostics diagnostics of of the the OPLS OPLS-DA-DA model, which gives an overview of the affecting variables on the model and clarifies model, which gives an overview of the affecting variables on the model and clarifies sig- significant metabolites in the projection (Figure3B). It shows that there were significantly nificant metabolites in the projection (Figure 3B). It shows that there were significantly higher metabolites in the uRBCs treated with AG, such as arginine, ribose, cis-aconitic acid, higher metabolites in the uRBCs treated with AG, such as arginine, ribose, cis-aconitic lactate, alanine, isovalerate, valine, acetone, glutamate, and riboflavin. These metabolites acid, lactate, alanine, isovalerate, valine, acetone, glutamate, and riboflavin. These metab- contributed to the separation in the upper-right quadrant of the S-plot, which have the olites contributed to the separation in the upper-right quadrant of the S-plot, which have covariances and positive correlations with uRBCs-AG. However, the other metabolites, the covariances and positive correlations with uRBCs-AG. However, the other metabo- such as mannose and 2-phosphoglycerta, were in the lower-left quadrant, which were lites, such as mannose and 2-phosphoglycerta, were in the lower-left quadrant, which higher in uRBCs, as illustrated in Figure3B. The results indicate that these metabolites were higher in uRBCs, as illustrated in Figure 3B. The results indicate that these metabo- were associated with AG’s effect, and they might be the biomarkers. lites were associated with AG’s effect, and they might be the biomarkers. In the other OPLS-DA model, an obvious distinction by PC1 between uRBCs and In the other OPLS-DA model, an obvious distinction by PC1 between uRBCs and uRBCs-CQ was shown in the OPLS-DA score plot with great discriminant statistical values uRBCsof R2- andCQ was Q2 representedshown in the as OPLS 0.903-DA and score 0.934, plot respectively. with great The discriminant uRBCs group statistical was on val- the uesleft of side R2 and of component Q2 represented 1, while as 0.903 uRBCs and treated 0.934, respectively. with CQ were The on uRBCs the right group side was of PC1on the(Figure left side4A). of These component results 1 are, while consistent uRBCs withtreated those with previously CQ were foundon the inright the side PCA of model. PC1 (FigureFigure 4A).4B represents These results the S-plot.are consistent There arewith substantially those previously higher found metabolites in the PCA in the model. group Figure 4B represents the S-plot. There are substantially higher metabolites in the group of uRBCs treated with CQ, such as arginine, lactate, glutamate, 2-aminobutyric, alanine, valine, isocitrate, acetone, threonine, proline, creatine, and lysine, which contribute to separation in the upper right side of the S-plot. At the same time, the bottom-left of the plot shows the metabolites which were low, such as 3-hydroxybutyrate and glutathione. Based on these findings, these metabolites are significantly affected by CQ and are possibly the biomarkers.

Metabolites 2021, 11, 486 8 of 24

of uRBCs treated with CQ, such as arginine, lactate, glutamate, 2-aminobutyric, alanine, valine, isocitrate, acetone, threonine, proline, creatine, and lysine, which contribute to separation in the upper right side of the S-plot. At the same time, the bottom-left of the plot shows the metabolites which were low, such as 3-hydroxybutyrate and glutathione. Metabolites 2021, 11, x FOR PEER REVIEW 9 of 26 Based on these ﬁndings, these metabolites are signiﬁcantly affected by CQ and are possibly the biomarkers.

Figure 4. The OPLS OPLS-DA-DA score score scatter scatter plot plot ( (AA)) and and S S-plot-plot ( (BB)) of of the the blood blood metabolic metabolic profiles proﬁles of of uninfected uninfected RBC RBC samples. samples. uRBCsuRBCs;; u uninfectedninfected red blood cells,cells, CQ;CQ; chloroquine.

TheThe variable variable importance importance of of projection projection (VIP) (VIP) showed showed that that 28 28 metabolites metabolites were were identi- iden- fiedtiﬁed as as biomarkers biomarkers from from the the OPLS OPLS-DA-DA model model of the of the uRBCs uRBCs treated treated with with AG. AG. The The VIP VIP > 1.0 > metabolites1.0 metabolites were were arginine, arginine, ribose, ribose, ribose, ribose, lactate, lactate, glutamate, glutamate, glutathione, glutathione, glutamate, glutamate, isocitrate,isocitrate, acetone, acetone, alanine, alanine, creatine, creatine, 2-aminobutyric 2-aminobutyric acid, acid, proline, proline, cis cis-aconitic-aconitic acid, acid, threo- threo- nine, riboflavin, ATP, valine, pyruvate, ornithine, phenylalanine, fructose, citrate, glucose, AMP, AMP, lysine, and glucose. (Figure 5). However, in the loading column plot, some of the metabolites were shown to be not significant due to their error bars crossing the 0 axis of the plot. Consequently, these metabolites did not have a significant impact on the separation.

Metabolites 2021, 11, 486 9 of 24

nine, riboflavin, ATP, valine, pyruvate, ornithine, phenylalanine, fructose, citrate, glucose, AMP, AMP, lysine, and glucose. (Figure5). However, in the loading column plot, some of the metabolites were shown to be not significant due to their error bars crossing the 0 Metabolites 2021,, 11,, xx FOFOR PEER REVIEW 10 of 26 axis of the plot. Consequently, these metabolites did not have a significant impact on the separation.

Figure 5.5. VariableVariable importance importance of of projection projection (VIP) (VIP values) values of of the the inﬂuential influential contributing contributing metabolites metabolites in the in OPLS-DAthe OPLS- scoreDA score plot ofplot uninfected of uninfected RBCs-control RBCs-control and uRBCs-CQ,and uRBCs- errorCQ,, errorerror bars representbarsbars represent the standard the standard error error of the of mean the mean (SEM), (SEM), uRBCs; uRBCs uninfected;; unin- redfected blood red cells,blood AG; cells andrographolide.,, AG;; andrographolide.

The VIP analysis of th thee second OPLS OPLS-DA-DA model was conducted to identify the most important contributing contributing variables variables,,, as as shown shown in in Figure Figure 6.6 .A Atotal total of of32 32metabolites metabolites signifi- sig- cantlynificantly contributed contributed to the to discrimination the discrimination between between the uRBCs the uRBCs treated treated with CQ with andCQ uRBCs and- controluRBCs-control with VIP with values VIP of values more than of more 1. The than metabolites 1. The metabolites were arginine, were lactate, arginine, glutamate, lactate, 2glutamate,-aminobutyric 2-aminobutyric acid, glutamate, acid, glutamate, alanine, ornithine, alanine, ornithine,valine, isocitrate, valine, isocitrate, acetone, arginine, acetone, threonine,arginine, threonine, proline, creatine, proline, creatine, cis-aconitic cis-aconitic acid, lysine, acid, ornithine, lysine, ornithine, fructose, fructose, phenylalanine, pheny- isobulalanine,tyrate, isobutyrate, AMP, AMP, AMP, ATP, AMP, riboflavin, ATP, riboflavin, glucose, glucose, pyruvate, pyruvate, glucose, glucose, ATP, lactate, ATP, lactate,valine, citrate,valine, and citrate, pyroglutamate. and pyroglutamate. These metabolites These metabolites had a significant had a significant impact on impactthe separation on the andseparation could be and suggested could be as suggested potential as biomarkers. potential biomarkers.

Figure 6. VIP values of the infl inﬂuentialuential contributing metabolites in the OPLS OPLS-DA-DA score plot of uninfected RBCs RBCs-control-control and uRBCs-CQ,uRBCs-CQ,, errorerror barsbars representrepresent the standardstandard error of the mean (SEM),(SEM),, uRBCs;uRBCs;; uninfecteduninfected red blood cells,cells,, CQ CQ;;; chloroquine.

2.4. The Model Validation of the OPLS-DA To validate the OPLS-DA model, in the internal cross-validation, if the values of R2Y and Q2 are higher than 0.5, this indicates the model is valid [38].. R2R2 denotesdenotes thethe fitnessfitness ofof the model as well as describing the quality of Y variables through the model, while Q2 indicates the predictive diagnostics of the model. In the present study, by calculating R2Y and Q2, the goodness of fit and the predicting ability were determined. From the OPLS-

Metabolites 2021, 11, 486 10 of 24

2.4. The Model Validation of the OPLS-DA To validate the OPLS-DA model, in the internal cross-validation, if the values of R2Y Metabolites 2021, 11, x FOR PEER REVIEWand Q2 are higher than 0.5, this indicates the model is valid [38]. R2 denotes the11 ﬁtness of 26

of the model as well as describing the quality of Y variables through the model, while Q2 indicates the predictive diagnostics of the model. In the present study, by calculating R2Y and Q2, the goodness of fit and the predicting ability were determined. From the DA model of uRBCs-AG and uRBCs, the model exhibited a high degree of fit, R2Y = 0.998, OPLS-DA model of uRBCs-AG and uRBCs, the model exhibited a high degree of fit, R2Y and a high degree of predictability, Q2 = 0.995. From the 100 permutations test, the model = 0.998, and a high degree of predictability, Q2 = 0.995. From the 100 permutations test, showed Y-axis intercepts of R2 less than 0.3 and those of Q2 less than 0.5 (R2 < 0.3, Q2 < the model showed Y-axis intercepts of R2 less than 0.3 and those of Q2 less than 0.5 (R2 0.05), indicating that the models are valid and did not show overfitting [38]. The permu- < 0.3, Q2 < 0.05), indicating that the models are valid and did not show overfitting [38]. tation tests with 100 permutations and external validation indicated that the OPLS-DA The permutation tests with 100 permutations and external validation indicated that the models were valid, as shown in Figure 7A. OPLS-DA models were valid, as shown in Figure7A.

Figure 7. The OPLS-DA models’ validation of uRBCs-control and uRBCs-AG: (A) 100 permutations test; (B) the validation Figure 7. The OPLS-DA models’ validation of uRBCs-control and uRBCs-AG: (A) 100 permutations test; (B) the validation regression plot. uRBCs, uninfected red blood cells; AG, andrographolide. regression plot. uRBCs, uninfected red blood cells; AG, andrographolide. Furthermore, this model is perceived to have excellent prediction and validation characteristics,Furthermore, which this model suggests is perceived that it is to an have adequate excellent model. prediction Additionally, and validation for the charac- estab- teristics,lished models, which suggests the validation that it is for an theadequate various model. variables Additionally, was verified for the using established the regression models, thediagnostics validation by for the the root various mean variables square error was verified of cross-validation using the regression (RMSECV) diagnostics and root by mean the rootsquare mean error square of calibration error of cross (RMSEC).-validation Based (RMSECV) on our results, and root the mean RMSEC square and error RMSECV of calibra- were tionnearly (RMSEC). similar, whichBased on implies our results, that it isthe a goodRMSEC model and (Figure RMSEC7B).V were These nearly results similar, suggest which that implieseach of that these it modelsis a good meet model the criteria(Figure of7B). great These validation results suggest and prediction that each performances. of these models meet Thethe criteria OPLS-DA of great model validation of uRBCs-CQ and prediction and uRBCs performances. gave the results of R2Y = 0.903 and Q2 =The 0.934, OPLS signifying-DA model that of this uRBCs model-CQ meets and theuRBCs criteria gave of the the results validation of R2Y and = prediction. 0.903 and Q2In addition,= 0.934, signifying the results that of the this 100 model permutations meets the test criteria were of acceptable, the validation which and indicates prediction. that Inthe addition, model is the valid results (Figure of 8theA). 100 The permutations regression coefficient test were (R acceptable,2) was 0.987. which Additionally, indicates that the thedifference model is between valid (Figure root mean 8A). squareThe regression error of cross-validationcoefficient (R2) was (RMSECV) 0.987. Additionally, and root mean the differencesquare error between of calibration root mean (RMSEC) square was error small. of cross These-validation findings show(RMSECV) that the and RMSEC root mean and squareRMSECV error are of nearly calibration similar, (RMSEC) which means was small. that itThese is a good findings model show (Figure that 8theB). RMSEC From these and RMSECVfindings, itare appears nearly thatsimilar, this modelwhich means has strong that validityit is a goo andd model prediction (Figure performances. 8B). From these findings, it appears that this model has strong validity and prediction performances.

Figure 8. The OPLS-DA model’s validation of uRBCs-control and uRBCs-CQ: (A) 100 permutations test; (B) the validation regression plot. uRBCs, uninfected red blood cells; CQ, chloroquine.

Metabolites 2021, 11, x FOR PEER REVIEW 11 of 26

DA model of uRBCs-AG and uRBCs, the model exhibited a high degree of fit, R2Y = 0.998, and a high degree of predictability, Q2 = 0.995. From the 100 permutations test, the model showed Y-axis intercepts of R2 less than 0.3 and those of Q2 less than 0.5 (R2 < 0.3, Q2 < 0.05), indicating that the models are valid and did not show overfitting [38]. The permutation tests with 100 permutations and external validation indicated that the OPLS-DA models were valid, as shown in Figure 7A.

Figure 7. The OPLS-DA models’ validation of uRBCs-control and uRBCs-AG: (A) 100 permutations test; (B) the validation regression plot. uRBCs, uninfected red blood cells; AG, andrographolide.

Furthermore, this model is perceived to have excellent prediction and validation characteristics, which suggests that it is an adequate model. Additionally, for the established models, the validation for the various variables was verified using the regression diagnostics by the root mean square error of cross-validation (RMSECV) and root mean square error of calibration (RMSEC). Based on our results, the RMSEC and RMSECV were nearly similar, which implies that it is a good model (Figure 7B). These results suggest that each of these models meet the criteria of great validation and prediction performances. The OPLS-DA model of uRBCs-CQ and uRBCs gave the results of R2Y = 0.903 and Q2 = 0.934, signifying that this model meets the criteria of the validation and prediction. In addition, the results of the 100 permutations test were acceptable, which indicates that the model is valid (Figure 8A). The regression coefficient (R2) was 0.987. Additionally, the difference between root mean square error of cross-validation (RMSECV) and root mean square error of calibration (RMSEC) was small. These findings show that the RMSEC and Metabolites 2021, 11, 486 RMSECV are nearly similar, which means that it is a good model (Figure 8B). From these11 of 24 findings, it appears that this model has strong validity and prediction performances.

Figure 8. The OPLS-DA model’s validation of uRBCs-control and uRBCs-CQ: (A) 100 permutations test; (B) the validation Figure 8. The OPLS-DA model’s validation of uRBCs-control and uRBCs-CQ: (A) 100 permutations test; (B) the validation regression plot. uRBCs, uninfected red blood cells; CQ, chloroquine. regression plot. uRBCs, uninfected red blood cells; CQ, chloroquine.

2.5. The Hierarchical Cluster Analysis (HCA) A hierarchical clustering analysis (HCA) was conducted to deduce the metabolites’ variations in different groups of the uRBCs. Clustering analysis is dependent on the simili- tude concept. One method to identify the similarity between two objects in mathematical terms is the Euclidean distance. The samples were divided into two groups. The first HCA was between the uRBCs-AG versus uRBCs and the second one was for uRBCs-CQ versus uRBCs. Then the visual HCA was performed to identify the metabolites’ discrepancies for each group. From the results of the first of OPLS-DA plots and VIP plots of uRBCs-AG and uRBCs, the metabolites that contributed the most to the discrimination were determined. Before being exposed to HCA with Euclidean distance measurements and Ward’s clustering algorithm, the distinctive binned regions of the metabolites of importance in the group were normalized and Pareto-scaled. Figure9 shows the results of the study as a HCA heat map, with each rectangle representing an averaged binned 1H-NMR spectral area indicative of the important metabolite, which was colored on a normalized scale from −1.5 (low) to 1.5 (high). The heat map successfully classified the group of uRBCs treated with AG as a large class that was different from uRBCs. In addition, the group of uRBCs treated with AG exhibited higher levels of arginine, ribose, lactate, glutamate, glutathione, isocitrate, acetone, alanine, creatine, 2-aminobutyric acid, proline, cis-aconitic acid, threonine, riboflavin, ATP, valine, pyruvate, ornithine, phenylalanine, fructose, citrate, AMP, and glucose. In contrast, these metabolites had lower levels in untreated uRBCs, as shown in Figure9. The results suggest that in the levels of the metabolites in the treated uRBCs group, exposure to AG significantly contributed to the changes observed in many of the metabolites in the uRBCs [41]. In addition to that, it is possible that these results are due to the fact that AG has the effect of compromising RBCs’ membrane integrity [30]. As mentioned earlier, a total of 28 metabolites were identified as biomarkers in the group of uRBCs-AG. The relative quantification of these particular biomarkers was then assessed through binned data of average peak metabolites within the groups. The variations in the level of the metabolites were quantitatively evaluated as shown in Figure 10, which represents the box plots of these significant individual metabolites. Metabolites 2021, 11, 486 12 of 24 Metabolites 2021, 11, x FOR PEER REVIEW 13 of 26

Figure 9. HeatFigure map 9. of Heat the identiﬁedmap of the biomarkers identified ofbiomarkers uRBCs and of uRBCs-AG uRBCs and obtained uRBCs- fromAG obtained HCA by from using HCA Ward’s by minimum variance methodusing and Ward’s Euclidean minimum distance. variance The method concentration and Euclidean of each metabolite distance. The is colored concentration based on of a each normalized metab- scale from minimum −olite1.5 (dark is colored blue) based to maximum on a no 1.5rmalized (dark brown).scale from uRBCs, minimum uninfected −1.5 (dark red blood blue) cells;to maximum AG, andrographolide. 1.5 (dark brown). uRBCs, uninfected red blood cells; AG, andrographolide.

Metabolites 2021, 11, 486 13 of 24 Metabolites 2021, 11, x FOR PEER REVIEW 14 of 26

Figure 10. Box plots of the relative quantities of the significant metabolites of uRBCs-AG (red color) and uRBCs-control Figure 10. Box plots of the relative quantities of the significant metabolites of uRBCs-AG (red color) and uRBCs-control 1 ≥ (green(green color) using 1HH-NMR-NMR spectra binned data with a VIP value ≥0.10.1 in in OPLS OPLS-DA-DA model. model. uRBCs, uRBCs, uninfected uninfected red red blood blood cells;cells; AG, AG, andrographolide. andrographolide. The other OPLS-DA plot and VIP plot of uRBCs-CQ and uRBCs in Figure 11 shows The other OPLS-DA plot and VIP plot of uRBCs-CQ and uRBCs in Figure 11 shows the results of the heat map analysis of identified biomarkers of uRBCs treated with CQ. The the results of the heat map analysis of identified biomarkers of uRBCs treated with CQ. uRBCs group treated with CQ was categorized successfully as a broad group that differed The uRBCs group treated with CQ was categorized successfully as a broad group that from uRBCs. Furthermore, there were higher levels of the metabolites of uRBCs treated differed from uRBCs. Furthermore, there were higher levels of the metabolites of uRBCs with CQ than in the uRBCs group. The metabolites were arginine, lactate, glutamate, treated with CQ than in the uRBCs group. The metabolites were arginine, lactate, gluta- 2-aminobutyric acid, glutamate, alanine, ornithine, valine, isocitrate, acetone, arginine, mate, 2-aminobutyric acid, glutamate, alanine, ornithine, valine, isocitrate, acetone, argi- threonine, proline, creatine, cis-aconitic acid, lysine, ornithine, fructose, phenylalanine, nine, threonine, proline, creatine, cis-aconitic acid, lysine, ornithine, fructose, phenylala- isobutyrate, AMP, AMP, ATP, riboflavin, glucose, pyruvate, glucose, ATP, lactate, valine, nine, isobutyrate, AMP, AMP, ATP, riboflavin, glucose, pyruvate, glucose, ATP, lactate, citrate, and pyroglutamate (Figure 11). As previously described, a total of 32 metabolites valine, citrate, and pyroglutamate (Figure 11). As previously described, a total of 32 me- in the uRBCs-CQ group were identified as biomarkers. The changes in metabolite lev- tabolitesels were in assessed the uRBCs via-CQ a comparison group were between identified uRBCs as biomarkers. treated with The CQchanges and thein metabolite untreated levelsgroup. wereFigure assessed 12 shows via a the comparison box plots ofbetween these important uRBCs treated individual with CQ metabolites, and the untreated relatively group.quantified Figure to TSP.12 shows the box plots of these important individual metabolites, relatively quantified to TSP.

Metabolites 2021, 11, 486 14 of 24 Metabolites 2021, 11, x FOR PEER REVIEW 15 of 26

Figure 11. HeatFigure map 11. of theHeat identiﬁed map of the biomarkers identified of biomarkers uRBCs-control of uRBCs and uRBCs-CQ-control and obtained uRBCs- fromCQ obtained HCA by from using Ward’s minimum varianceHCA by method using and Ward’s Euclidean minimum distance. variance The concentration method and ofEuclidean each metabolite distance. is coloredThe concentration based on a normalizedof scale from minimumeach metabolite−1 (dark is blue) colored to maximumbased on a 1 normalized (dark brown). scale uRBCs, from uninfectedminimum −1 red (dark blood blue) cells; to CQ, maximum chloroquine. 1 (dark brown). uRBCs, uninfected red blood cells; CQ, chloroquine.

Metabolites 2021, 11, 486 15 of 24 Metabolites 2021, 11, x FOR PEER REVIEW 16 of 26

Figure 12.Figure Box 12.plotsBox of plotsthe relative of the relativequantities quantities of the significant of the signiﬁcant metabolites metabolites of uRBCs of-CQ uRBCs-CQ (green color) (green and color) uRBCs and (red uRBCs (red 1 color) usingcolor) H using-NMR1 H-NMRspectra binned spectra data binned with data VIP with value VIP ≥0.1 value in OPLS≥0.1- inDA OPLS-DA model. uRBCs model., uninfected uRBCs, uninfected red blood red cells; blood CQ, cells; CQ, chloroquine. chloroquine.

2.6. Analysis2.6. Analysis of the Perturbed of the Perturbed Metabolic Metabolic Pathways Pathways by AG and by AGCQ andin Uninfected CQ in Uninfected RBCs RBCs The identifiedThe identiﬁed metabolite metabolite perturbations perturbations based on based the 1 onH-NMR the 1H-NMR data are data useful are for useful ex- for ex- tractingtracting out the outdesired the desired information information from the from uRBCs the uRBCs and uRBCs and uRBCs treated treated groups. groups. To sys- To system- tematicallyatically identify identify the themost most significant signiﬁcant pathways pathways that that are are involved involved in these in these groups, groups, met- metabolic abolic pathway analysis (MetPA) using MetaboAnalyst

Metabolites 2021, 11, x FOR PEER REVIEW 17 of 26

(www.metaboanalyst.ca/MataboAnalyst) accessed on 20 March 2021 was performed as well as KEGG (https://www.genome.jp/kegg/pathway) accessed on 1 March 2021. The pathway impact factors are the appropriate tool to measure the significance of metabolites in the network. From the results, 28 metabolic pathways were identified in the group of Metabolites 2021, 11, 486 uRBCs-AG (Table S1). Based on the set standards for a pathway impact value of16 ofhigher 24 than 0.1 [40] from the 28 metabolic pathways, 10 metabolic pathways were determined as important metabolic pathways in the uRBCs which were exposed to AG. The D-glutamate and D-glutamine metabolism as well as metabolic of riboflavin showed the same highest pathway analysis (MetPA) using MetaboAnalyst (www.metaboanalyst.ca/MataboAnalyst) impactaccessed value onof 0.50, 20 March followed 2021 wasby phenylalanine performed as well metabolism, as KEGG (argininehttps://www.genome.jp/ and proline metabolism,kegg/pathway glutathione )metabolism, accessed on arginine 1 March 2021.biosynthesis, The pathway citrate impact cycle, factors pyruvate are themetabolism, appro- alanine,priate aspartate tool to measureand glutamate the significance metabolism, of metabolites and glycolysis/gluconeogenesis in the network. From the with results, values of 0.35,28 metabolic0.34, 0.27,0.25, pathways 0.23, were 0.20, identified 0.19, and in 1.0, the respectively group of uRBCs-AG (Table 2 (Tableand Figure S1). Based 13). on the set standards for a pathway impact value of higher than 0.1 [40] from the 28 metabolic Table 2. Results of the ingenuity pathways,pathway analysis 10 metabolic of uninfected pathways RBCs were treated determined with AG as by important using MetaboAnalyst metabolic pathways (MetPA) in shown the impact values. the uRBCs which were exposed to AG. The D-glutamate and D-glutamine metabolism as well as metabolic of riboflavin showed the same highest impact value of 0.50, followed NO. Pathway Name by phenylalanineMatch metabolism, Raw argininep and-log(p) proline metabolism,Holm p glutathioneFDR metabolism,Impact 1. arginine biosynthesis, citrate cycle, pyruvate metabolism, alanine, aspartate and glutamate Arginine and proline metabolism 6/38 0.0000141 4.8507 0.0011704 0.0005922 0.34441 metabolism, and glycolysis/gluconeogenesis with values of 0.35, 0.34, 0.27,0.25, 0.23, 0.20, 2. Citrate cycle (TCA cycle) 0.19, and 1.0, respectively4/20 (Table0.00018192 and Figure3.7401 13 ). 0.014736 0.0038204 0.23173 3. Alanine,Table aspartate 2. Results and of the glutamate ingenuity pathwaymetabolism analysis 4/28 of uninfected0.0007073 RBCs treated3.1504 with AG by0.056585 using MetaboAnalyst 0.011883 (MetPA) 0.19712 4. shown the impact values. Arginine biosynthesis 3/14 0.0010628 2.9735 0.083964 0.01488 0.2538 NO. Pathway Name Match Raw p -log(p) Holm p FDR Impact 5. Glycolysis/Gluconeogenesis 3/26 0.0067171 2.1728 0.51721 0.070529 0.10065 1. Arginine and proline metabolism 6/38 0.0000141 4.8507 0.0011704 0.0005922 0.34441 6. 2. Citrate cycle (TCA cycle) 4/20 0.0001819 3.7401 0.014736 0.0038204 0.23173 Glutathione metabolism 3/28 0.0082928 2.0813 0.63026 0.0774 0.27562 Alanine, aspartate and glutamate 3. 4/28 0.0007073 3.1504 0.056585 0.011883 0.19712 7. metabolism Pyruvate metabolism 2/22 0.043961 1.3569 1.0 0.3357 0.20684 4. Arginine biosynthesis 3/14 0.0010628 2.9735 0.083964 0.01488 0.2538 8. 5. Glycolysis/Gluconeogenesis 3/26 0.0067171 2.1728 0.51721 0.070529 0.10065 Riboflavin metabolism 1/4 0.060569 1.2177 1.0 0.39137 0.5 6. Glutathione metabolism 3/28 0.0082928 2.0813 0.63026 0.0774 0.27562 9. 7. Pyruvate metabolism 2/22 0.043961 1.3569 1.0 0.3357 0.20684 D-Glutamine and D-glutamate metabolism 1/6 0.089519 1.0481 1.0 0.47604 0.5 8. Riboflavin metabolism 1/4 0.060569 1.2177 1.0 0.39137 0.5 10. D-Glutamine and D-glutamate Phenylalanine9. metabolism 1/61/10 0.14488 0.089519 0.839 1.0481 1.01.0 0.476040.67609 0.5 0.35714 metabolism Raw10. p-values Phenylalanine were defined metabolism according to a total number 1/10 of hits 0.14488 and total 0.839compounds 1.0 in each pathway; 0.67609 Holm p, 0.35714 p-value correctedRaw byp -valuesHolm– wereBonferroni defined accordingmethod, toFDR, a total false number discovery of hits andrate; total impact, compounds the pathway in each pathway; impact Holmvalue p, computedp-value corrected from bypath- way topologyHolm–Bonferroni analysis. method, FDR, false discovery rate; impact, the pathway impact value computed from pathway topology analysis.

FigureFigure 13. Summary 13. Summary of the of thepathway pathway analysis analysis of of uninfected uninfected RBCsRBCs treated treated with with AG. AG.

Metabolites 2021, 11, x FOR PEER REVIEW 18 of 26

However, in the uRBCs-CQ group, the number of the identified metabolic pathways, including significant pathways, was less than in the uRBCs-AG group. Twenty-six metabolic pathways were identified in this group (Table S2). From 26 metabolic pathways, nine metabolic pathways were identified as important metabolic pathways in uRBCs treated with CQ. In addition, all affected metabolic pathways in this group were the same path- Metabolites 2021, 11, 486 ways identified in the uRBCs-AG group, except for glutathione metabolism, which17 of 24 was not significant because its impact value was less than 1.0. This is illustrated in Table 3 and Figure 14. Further, the general metabolic pathway map was constructed to illustrate all biomarkersHowever, and roles in the of uRBCs-CQeach metabolite group, in the the number different of thepathways identified of the metabolic treated path- group of uRBCsways,-AG including and uRBCs significant-CQ. pathways, All these was metabolites less than in took the uRBCs-AG part in a numbergroup. Twenty-six of pathways, whichmetabolic include pathways amino acid, were carbohydrate, identified in this and group lipid (Table metabolism. S2). From This 26 schematic metabolic path-metabolic pathwayways, nine is illustrated metabolic pathwaysin Figure were15. identified as important metabolic pathways in uRBCs treated with CQ. In addition, all affected metabolic pathways in this group were the same pathways identified in the uRBCs-AG group, except for glutathione metabolism, which Table 3. Results of the ingenuity pathway analysis of uninfected RBCs treated with CQ by using MetaboAnalyst (MetPA) was not significant because its impact value was less than 1.0. This is illustrated in Table3 showing the impact values. and Figure 14. Further, the general metabolic pathway map was constructed to illustrate NO. Pathway Name all biomarkers and rolesMatch of each Raw metabolite p -log(p) in the differentHolm pathways p FDR of the treated groupImpact 1. of uRBCs-AG and uRBCs-CQ. All these metabolites took part in a number of pathways, Arginine and proline metabolismwhich include amino6/38 acid, carbohydrate,0.000014 4.8507 and lipid metabolism.0.001704 This0.00059 schematic metabolic0.34 pathway is illustrated in Figure 15. 2. Citrate cycle (TCA cycle) 4/20 0.000181 3.7401 0.014736 0.00304 0.23

3. Alanine,Table 3. aspartateResults of and the ingenuityglutamate pathway metabolism analysis 4/28 of uninfected 0.000707 RBCs treated3.1504 with CQ0.056585 by using MetaboAnalyst0.0113 (MetPA)0.20 showing the impact values.

4. Arginine biosynthesis 3/14 0.00128 2.9735 0.083964 0.018 0.25 NO. Pathway Name Match Raw p -log(p) Holm p FDR Impact 5. 1.Glycolysis/Gluconeogenesis Arginine and proline metabolism 6/383/26 0.0000140.00671 4.85072.1728 0.0017040.51721 0.000590.0729 0.34 0.10 6. 2. Citrate cycle (TCA cycle) 4/20 0.000181 3.7401 0.014736 0.00304 0.23 PyruvateAlanine, metabolism aspartate and glutamate 2/22 0.043961 1.3569 1.0 0.335 0.21 3. 4/28 0.000707 3.1504 0.056585 0.0113 0.20 metabolism 7. 4.Riboflavin Arginine metabolism biosynthesis 3/141/4 0.001280.060569 2.97351.2177 0.0839641.0 0.0180.39137 0.25 0.50 5. Glycolysis/Gluconeogenesis 3/26 0.00671 2.1728 0.51721 0.0729 0.10 8. 6.D-Glutamine Pyruvate and metabolism D-glutamate metabolism 2/221/6 0.0439610.089519 1.35691.0481 1.01.0 0.3350.47604 0.21 0.50 7. Riboﬂavin metabolism 1/4 0.060569 1.2177 1.0 0.39137 0.50 9. PhenylalanineD-Glutamine metabolism and D-glutamate 1/10 0.14488 0.839 1.0 0.67609 0.36 8. 1/6 0.089519 1.0481 1.0 0.47604 0.50 metabolism Raw9. p-values Phenylalanine were defined metabolism according to a total 1/10 number of 0.14488 hits and total 0.839 compounds 1.0 in each pathway 0.67609. Holm 0.36p, p-value corrected by Holm–Bonferroni method, FDR, false discovery rate; impact, the pathway impact value computed from path- Raw p-values were deﬁned according to a total number of hits and total compounds in each pathway. Holm p, p-value corrected by way topologyHolm–Bonferroni analysis. method, FDR, false discovery rate; impact, the pathway impact value computed from pathway topology analysis.

FigureFigure 14. 14. SummarySummary of of the the pathway pathway analysis of of uninfected uninfected RBCs RBCs treated treated with with CQ. CQ.

Metabolites 2021, 11, 486 18 of 24 Metabolites 2021, 11, x FOR PEER REVIEW 19 of 26

Figure 15. Schematic representation showing the perturbated metabolic pathways of the uninfected RBCs exposed to AG (uRBCs-AG)(uRBCs-AG) and the uninfected RBCs exposed to CQ (uRBCs-CQ)(uRBCs-CQ) as identifiedidentified by 11H-NMR.H-NMR. The metabolites that are outlined with light blue are present in uRBCs-AG,uRBCs-AG, red are presents in uRBCs-CQ,uRBCs-CQ, whereas those highlighted in gold are present in both treated groups, uRBCs-CQ and uRBCs-AG. The metabolic pathways include, (1) glutathione metabolism; present in both treated groups, uRBCs-CQ and uRBCs-AG. The metabolic pathways include, (1) glutathione metabolism; (2) (2) arginine biosynthesis; (3) riboflavin metabolism; (4) glycolysis/c; (5) citrate cycle/TCA cycle; (6) pyruvate metabolism; arginine biosynthesis; (3) riboflavin metabolism; (4) glycolysis/c; (5) citrate cycle/TCA cycle; (6) pyruvate metabolism; (7) (7) alanine, aspartate, and glutamate metabolism; (8) D-glutamine and D-glutamate metabolism; (9) phenylalanine metab- olismalanine,; and aspartate, (10) arginine and glutamate and proline metabolism; metabolism. (8) D-glutamine and D-glutamate metabolism; (9) phenylalanine metabolism; and (10) arginine and proline metabolism. From the results of PAC and OPLS-DA, both treated groups of uRBCs-AG and uR- From the results of PAC and OPLS-DA, both treated groups of uRBCs-AG and uRBCs- BCs-CQ exhibited distinct variations in their metabolites compared to the uRBCs-control CQ exhibited distinct variations in their metabolites compared to the uRBCs-control group. group. From the analysis of the perturbated metabolic pathways, ten metabolic pathways From the analysis of the perturbated metabolic pathways, ten metabolic pathways in in uninfected RBCs treated with AG have been determined as disturbed. On the other uninfected RBCs treated with AG have been determined as disturbed. On the other hand, hand, nine metabolic pathways in uRBCs treated with CQ were identified as disrupted nine metabolic pathways in uRBCs treated with CQ were identified as disrupted metabolic metabolic pathways. In this section, the suggested metabolic pathways are further dis- pathways. In this section, the suggested metabolic pathways are further discussed in detail. cussed in detail. 2.6.1. Amino Acid Metabolism 2.6.1. Amino Acid Metabolism In the human body, RBCs are the most abundant host cells. They represent a large proportionIn the human of the free body, amino RBCs acids are(A.A.) the most in the abundant blood [ 42host–44 cells.], due They to the represent fact that maturea large proportionRBCs do not of havethe free nuclei, amino mitochondria, acids (A.A.) ribosomes,in the blood and [42 other–44], due organelles to the fact and that are matur unablee RBCsto synthesize do not have new proteinsnuclei, mitochondria, [4,5]. Nevertheless, ribosomes, various and A.A other transport organelles systems and are have unable been toidentified synthesize in humannew proteins RBCs, [4,5] which. Nevertheless, permit the uptake various of A.A various transport exogenous systems biochemical have been identifiedcompounds, in human such as RBCs, amino which acids, carbohydrates,permit the uptake and of inorganic various ionsexogenous [45] or endogenousbiochemical compounds, such by the as RBCs amino [34 acids,]. In addition,carbohydrates, the transport and inorganic system ions resembles [45] or those endogenous in other compounds,cells [46]. The by results the RBCs of the [34] present. In addition, study show the transport an increase system in the resembles concentrations those in of other most cellsA.A. [46] (such. The as arginine,results of alanine,the present threonine, study sh valine,ow an ornithine, increase in glutamate, the concentrations lysine, phenylala- of most A.A.nine, (such methionine, as arginine, glutamine, alanine, and threonine, proline) in valine, the uRBCs ornithine, exposed glutamate, to CQ. The lysine, increased phenylala- level nine,of A.A. methionine, in this group glutamine, suggests thatand theproline) chloroquine in the uRBCs might inhibit exposed the to RBCs CQ. in The the increased uptake of levelsome of amino A.A.acids, in this thus group increasing suggests their that levels the chloroquine [47]. Furthermore, might theinhibit concentrations the RBCs in of the uptakeA.A. mentioned of some amino above acids, were observedthus increasing to rise their in the levels uRBCs [47] exposed. Furthermore, to AG. The the highconcentra- levels tioofns these of the A.A. A.A. in thismentioned group can above be considered were observed an early to rise sign in thatthe uRBCs the mechanism exposed ofto actionAG. The of highAG is levels associated of these with A.A. levels in this of group A.A. The can resultsbe considered of HCA an showed early sign that that there the is mechanism an evident ofrelationship action of AG between is associated AG and with CQ associatedlevels of A.A. with The a concentration results of HCA of A.A.showed Therefore, that there it can is anbe suggested evident relationship that AG and between CQ may AG have and the CQ effect associated of increasing with the a concentration levels of A.A. of of RBCs. A.A.

Metabolites 2021, 11, 486 19 of 24

2.6.2. Glutathione Metabolism The current study results reveal a signiﬁcant increase in the level of GSH in the uRBCs exposed to AG compared to the group of uRBCs exposed to CQ. Glutathione (GSH) is a tripeptide of low molecular weight, and it is well-known that GSH performs a crucial protective role versus oxidative processes, which damages the hemoglobin function and results in RBCs’ lysis [48,49]. In order to evaluate redox status, the antioxidants GSH and GSSG were previously determined and characterized by 1H-NMR in healthy RBCs [50–52]. The results of the current study reveal the differences in the levels of GSH in the treated groups, a result which is in alignment with Rossi et al.’s (2002) [53]. AG is one of the bioactive compounds of A. paniculate as a labdane diterpenoid derivative. Several publications have revealed the biological properties of AG, such as antimicrobial, anti-inﬂammation, and antioxidant [19,54]. The increase of GSH level can possibly be associated with the antioxidative property of AG. Moreover, these increases may be as a result of a conversion of GSSG to GSH [41]. It is known that the synthesis of GSH requires three amino acids, viz. glutamate, cysteine, and glycine, as well as two enzymes, glutamate-cysteine ligase (GCL) and GSH synthetase, which require ATP [48,55]. From the results, these A.A. were determined to be present at a high concentration, which in turn contributes to the increased synthesis of GSH.

2.6.3. Carbohydrate Metabolism The results showed high levels of glucose, ribose, and fructose in the uRBCs exposed to AG group compared to those in the uRBCs group. Meanwhile, the levels of glucose and fructose metabolites were high in the uRBCs exposed to CQ group. Such ﬁndings suggest that AG and CQ might affect the pathways associated with carbohydrate metabolism. The metabolism of carbohydrates is the most essential metabolic pathway, including glycolysis and gluconeogenesis, pyruvate metabolism and citrate cycle (TCA), due to its role in controlling and modulating the production of cellular energy in organisms. For example, the catabolism of glucose produces pyruvate with its derived ATP and NADH via the glycolysis pathway, which usually needs aerobic conditions for the conversion process to the acetyl-CoA essential for the TCA cycle [56]. In anaerobic conditions, glucose produces pyruvate, after which it is reduced to lactate [57]. As mentioned earlier, the metabolic pathways were identiﬁed as metabolic disturbances after exposure of the uRBCs to AG. These disorders in uRBCs may be attributed to AG (Table2). Moreover, the results showed a notable increase in the level of lactate in both treated groups, uRBCs-CQ and uRBCs-AG, respectively, compared to the uRBCs group (Figures9 and 11). This is probably because of the increased lactate metabolism and the change in energy metabolism from glucose to lactate in the glycolysis pathway [57].

3. Materials and Methods 3.1. Chemicals and Consumables Human O- erythrocytes, Roswell Park Memorial Institute medium (RPMI-1640) medium (1X) with sodium bicarbonate, Albumax II was procured from Gibco BRL (Grand Island, NY, USA). HEPES, triton X-100, EDTA, saponin, sorbitol, hypoxanthine, (100X) phosphate buffered saline (PBS), chloroquine diphosphate (CQ), NMR solvent: deuterium oxide 99.9% (D2O), trimethylsilyl-2,2,3,3-tetradeuteropropionate acid sodium salt (TSP), potassium hydroxide (KOH), and potassium dihydrogen phosphate (KH2PO4) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Andrographolide compound C20H30O5 (Cat. No.: M16-11074), presented in Figure 16, was also obtained from Sigma-Aldrich. Gen- tamicin, sodium deuteroxide solution (NaOD), sodium chloride (NaCl) and hydro-chloric acid (HCl) were purchased from Jiangxi Dongxu Chemical Technology Co., Ltd, Kuala Lampur, Malaysia. Metabolites 2021, 11, x FOR PEER REVIEW 21 of 26

Metabolites 2021, 11, 486 20 of 24 chloric acid (HCl) were purchased from Jiangxi Dongxu Chemical Technology Co., Ltd, Kuala Lampur, Malaysia.

FigureFigure 16.16. TheThe chemicalchemical structurestructure ofof thethe andrographolideandrographolide compound.compound.

3.2.3.2. BloodBlood CollectionCollection andand PreparationPreparation forfor CultureCulture TheThe firstfirst authorauthor donated donated uninfected uninfected RBCs RBCs (type (type O- O negative)- negative) under under the the supervision supervision of aof hematologist. a hematologist. The The blood blood was was combined combined with with citrate citrate phosphate phosphate buffer buffer as an as anticoagulant an anticoag- (1:9ulant anticoagulant/blood). (1:9 anticoagulant/blood). The blood The wasblood then was washed then washed three times three with times washing with medium washing tomedium remove to plasma remove and plasma white and blood white cells blood before cells being before resuspended being resuspended to generate to generate an RBC suspension.an RBC suspension. In the washing In the mediawashing with media RPMI-1640, with RPMI the- following1640, the combinationfollowing combination was used, includingwas used, 25 including mM HEPES 25 (4-(2-hydroxyethyl)-1-piperazine-ethan-sulphonicmM HEPES (4-(2-hydroxyethyl)-1-piperazine-ethan acid)-sulphonic buffer (pHacid) 7.4), buffer 24 mM(pH sodium7.4), 24 mM bicarbonate, sodium bicarbonate, 11 mM glucose, 11 mM and glucose, 50 g/L gentamicin.and 50 g/L gentamicin. The usual bloodThe usual washing blood process washing was process used as was previously used as previously described [described53]. [53]. 3.3. In-Vitro Cell Culture of Uninfected RBCs 3.3. In-Vitro Cell Culture of Uninfected RBCs The uRBCs were cultured as described by Trager and Jensen [58] and suspended in completeThe RPMI-1640uRBCs were malaria cultured culture as described medium by (cMCM) Trager containing and Jensen 25 [58] mM and HEPES, suspended 0.75 mM in hypoxanthin,complete RPMI Albumax-1640 malaria 5%, 24 culture mM sodium medium bicarbonate, (cMCM) containing 11 mM glucose, 25 mM and HEPES, 20 µg/mL 0.75 gentamicinmM hypoxanthin, at pH and Albumax hematocrit 5%, levels 24 mM of 7.4 sodium and 2%, bicarbonate, respectively, 11 in mM 75 cm glucose,2 tissue culture and 20 2 flasks.µg/mL After gentamicin that, all at cultures pH and were hematocrit incubated levels at standardof 7.4 and conditions 2%, respectively, for parasite in 75 cultivation cm tissue culture flasks. After that, all cultures were incubated at standard◦ conditions for parasite in a micro-aerophilic environment containing 5% CO2 at 37 C. The medium was changed daily.cultivation Before in the a micro drug exposure-aerophilic commenced, environment the containing uRBCs were 5% established CO2 at 37 at°C. 2% The hematocrit. medium Thewas firstchanged group dai wasly. Before the untreated the drug uRBCs exposure cultures commenced, flasks which the uRBCs served were as a established control. The at second2% hematocrit. group was The uRBCs first group treated was by the exposure untreated to AG uRBCs at a concentration cultures flasks of which 4.14 µ M,served which as a control. The second group was uRBCs treated by exposure to AG at a concentration of represented the IC50 of AG (the concentration of AG that inhibits malaria parasite growth at4.14 50%), µM, and whic theh represented last group was the IC uRBCs50 of AG treated (the concentration by exposure to of CQ AG at that a concentration inhibits malaria of 20.19parasite nM growth (the concentration at 50%), and of the CQ last that group inhibits was malaria uRBCs parasite treated growth by exposure at 50%), to theCQ ICat50 a ofconcentration both AG and of CQ20.19 were nM obtained(the concentration from a previous of CQ that study inhibits [59]. Next,malaria the parasite samples growth were collectedat 50%), the from IC the50 of control both AG and and uRBCs-CQ CQ were after obtained 12 h of from drug a exposure. previous Thestudy uRBCs-AG [59]. Next, were the collectedsamples were after 24collected h; 12 and from 24 the h were control the and initial uRBCs times-CQ of action after 12 of h CQ of anddrug AG exposure. against The the P.uRBCs falciparum-AG were3D7, collected which was after determined 24 h; 12 and in 24 a previous h were the study initial [59 times], respectively. of action of For CQ each and group,AG against five culturethe P. falciparum flasks were 3D7, prepared, which was and determined the procedure in wasa previous repeated study three [59] times., respectively. For each group, five culture flasks were prepared, and the procedure was repeated 3.4.three The times. Extraction of Metabolites of Uninfected RBC The uRBCs were extracted as described by Pertinhez et al. [16]. After 12 h and 24 h of3.4. drug The Extraction exposure, of the Metabolites supernatant of Uninfected was removed RBC from the cultures after centrifugation ◦ at 2000The rpm uRBCs for 5were min extracted at 4 C. Then,as described the pellets by Pertinhez of untreated et al. uRBCs [16]. After as well 12 h as and treated 24 h uRBCsof drug wereexposure, individually the supernatant washed was in the removed 0.9% NaCl/10 from the mMcultures phosphate after centrifugation buffer (pH 7).at Following2000 rpm for this 5 min procedure, at 4 °C. theThen, supernatant the pellets wasof untreated discarded uRBCs after as centrifugation well as treated as uRBCs stated earlier.were individually This step was washed repeated in the twice. 0.9% Subsequently,NaCl/10 mM phosphate the samples buffer were (pH lysed 7). throughFollowing a ◦ ◦ cyclethis procedure, of freezing the at −supernatant80 C, thawed was atdiscard 37 C,ed vortexed after centrifugation for 20 s, and as finally stated sonicated earlier. This for 30step s. was The repeated procedure twice. was Subsequently, repeated three the times. samples Next, were the lysed proteins through and membranesa cycle of freezing of all samplesat −80 °C, were thawed excluded at 37 °C, by ultrafiltrationvortexed for 20 using s, and a 5finally kDa cut-offsonicated filter for [60 30]. s. All The the procedure samples were centrifugated through a 5 kDa cutoff, and filtered at 14,000 rpm for 30 min at 4 ◦C[16]. Subsequently, 400 µL from each sample was transferred to an Eppendorf tube and mixed with 200 µL of phosphate buffer solution (KH2PO4) containing 0.1% trimethylsily l-2,2,3,3-

Metabolites 2021, 11, 486 21 of 24

tetradeuteropropionate acid sodium salt (TSP) in D2O (pH~7.4). A volume of 550 µL from each solution was transferred into a 5mm NMR tube. The NMR tubes were marked and subjected to 1H-NMR.

3.5. H-NMR Spectroscopy Analysis All NMR spectra were acquired at 499.887 MHz on a 500 MHz Varian INOVA NMR spectrometer (Varian Inc., California, USA). The spectral analysis was carried out at room temperature with presaturation setting to suppress the residual water signal. Then, the transverse relaxation time of T2 measurement CPMG (Carr–Purcell–Meiboom–Gill) experiment was performed using the following parameters: (δ = 0.0002); (big δ = 0.4); a number of transient (n = 256 scans), and relaxation delay RD (D1 = 0.5 s). The CPMG spectrum was obtained in 16 min and 1 s. The CPMG experiment can reduce the broad signals of macromolecules and minimize the intensity to obtain a much better spectral baseline [61]. The TSP was used as the internal reference for the calibration of chemical shift at δ 0.0 ppm.

3.6. Processing Data of 1H-NMR Spectra and Metabolite Identification One-dimensional (1D) 1H-NMR spectra were achieved on 500 MHz NMR spectrometer. All NMR spectra were processed by phased, baseline corrected, and calibrated to the internal standard (TSP) at 0.00 ppm. The NMR spectra were converted to ASCII files using Chenomx NMR suite software 7.5 (Chenomx Inc., Edmonton, Alberta, Canada). The NMR spectra were binned into 245 bins in the width of 0.04 ppm with the omission of the water region from (δ 4.75–4.90 ppm). MestReNova 6.0.2 software (Mestrelab Research, Santiago de Compostela, Spain) was used to process and assist for biomarker identification. The metabolites were identified by matching their signals with the library of Chenomx NMR database (version 7.5), recent publications on human blood and human serum metabolome [2,17], and the Human Metabolome Database (HMDB); http://www.hmdb.ca accessed on 14 March 2021 [62]. Afterwards, the NMR data were imported to the SIMCA-P software package (version 14.1) accessed on 14 March 2021. Pareto scaling was used to measure the data in order to reduce the effects of noise as well as make all metabolite signals the same strength [38].

3.7. Statistical Analysis For the analysis and determination of variation between untreated and treated uRBCs, multivariate data analysis MVDA was used using the SIMCA-P software package (version 14.1). The data were analyzed and visualized by using an unsupervised statistical principal component analysis (PCA) and supervised statistical orthogonal partial least squares- discriminant analysis (OPLS-DA). In addition, the validation and signiﬁcance of the model were carried out using the SIMCA-P software, 100 permutation tests, CV-ANOVA test, regression line, as well as the computation of R2Y and Q2Y values [38,40]. The heat map for the correlation analysis metabolites was carried out using MetaboAnalyst 4.0 (http://www.metaboanalyst.ca) accessed on 20 March 2021 [40]. The variable importance in projection (VIP) values of >1 [63] were used to determine the potential biomarkers in OPLS-DA. The pathway analyses were performed using the MetaboAnalyst 4.0 together with the Kyoto encyclopedia of genes and genomes (KEGG), and a database (http://www. genome.jp/kegg/ExPASy) accessed on 1 March 2021. The one-way analysis of variance (ANOVA) was performed using MetaboAnalyst. Tukey’s test was chosen as the post hoc analysis method and p ≤ 0.05 was considered to be statistically signiﬁcant, while the values were expressed as mean ± SEM.

4. Conclusions In conclusion, the experiments presented in this study were undertaken to determine the pharmacological effects of andrographolide as well as chloroquine on the metabolic variations of uninfected red blood cells (uRBCs) in vitro using a 1H-NMR-based metabolomics approach in combination with MVDA. Based on S-plot and VIP values, a total of 28 and Metabolites 2021, 11, 486 22 of 24

32 metabolites were identified as biomarkers in uRBCs-AG and uRBCs-CQ, respectively. In uRBCs treated with AG, ten metabolic pathways were determined as affected metabolic pathways, including riboflavin, D-glutamine and D-glutamate, phenylalanine, arginine and proline, glutathione, pyruvate alanine, aspartate, and glutamate metabolisms as well as arginine biosynthesis, citrate cycle, and glycolysis/gluconeogenesis. In contrast, in CQ-treated uRBCs, nine metabolic pathways were determined as disturbed metabolic pathways, which were the same disturbed metabolic pathways mentioned above for uRBCs-AG, except for glutathione metabolism. These findings suggest an evident relationship between AG and CQ associated with metabolic perturbations in intact RBCs after being exposed to the treatment. Our metabolomics results could allow valuable, comprehensive insights into the underlying mechanism of action of AG and CQ on red blood cells.

Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/metabo11080486/s1, Table S1: The ingenuity pathway analysis of uRBCs-AG by using Metabo- Analyst (MetPA), Table S2: The ingenuity pathway analysis of uRBCs-CQ using MetaboAnalyst (MetPA). Author Contributions: Conceptualization, A.A.I.A., Z.O.I., R.A.M. and I.S.I.; data curation, A.A.I.A.; formal analysis, A.A.I.A., A.M., R.A.M. and I.S.I.; funding acquisition, R.A.M.; investigation, A.A.I.A.; methodology, A.A.I.A., Z.O.I., N.Z.U., R.A.M., S.A.A., M.N.W. and R.B.; project administration, R.A.M. and I.S.I.; resources, N.N., N.Z.U., R.A.M., S.A.A., M.N.W., R.B. and I.S.I.; software, A.A.I.A., A.M. and I.S.I.; supervision, N.N., R.A.M., R.B. and I.S.I.; validation, A.A.I.A. and A.M.; writing— original draft, A.A.I.A., A.M., R.B. and I.S.I.; writing—review and editing, A.A.I.A., A.M. and R.B. All authors have read and agreed to the published version of the manuscript. Funding: This research received funding from the Ministry of Higher Education, Malaysia, by supporting this work through the research grant number: 5524777, University Putra Malaysia. Institutional Review Board Statement: Ethical review and approval were waived for this study due to (the blood sample was collected from the first author, and the IACUC has ruled out that no ethical consent is required for blood sampling taken from oneself by the person itself). Informed Consent Statement: Patient consent was waived due to the blood being collected from the first author. Data Availability Statement: The data used to support the findings of this study are included within the article and the Supplementary Information files. Any other data can be made available upon request. Acknowledgments: The first author gratefully acknowledges support from the Ministry of Higher Education Libya for the scholarship. The authors acknowledge the Faculty of Medicine and Health Sciences, University Putra Malaysia for providing the required facilities and equipment to run the experiment. We thank Nurkhalida Binti Kamal from Institute of Systems Biology (INBIOSIS), Universiti Kebangsaan Malaysia, UKM Bangi, Selangor, Malaysia, for her valuable comments and suggestions that improved the quality of this manuscript. Conflicts of Interest: The authors declare that there is no conflict of interest regarding this manuscript.

References 1. Pierigè, F.; Seraﬁni, S.; Rossi, L.; Magnani, M. Cell-based drug delivery. Adv. Drug Deliv. Rev. 2008, 60, 286–295. [CrossRef] 2. Nagana Gowda, G.A.; Raftery, D. Whole Blood Metabolomics by 1H NMR Spectroscopy Provides a New Opportunity to Evaluate Coenzymes and Antioxidants. Anal. Chem. 2017, 89, 4620–4627. [CrossRef] 3. Di Pierro, D.; Ciaccio, C.; Sbardella, D.; Tundo, G.R.; Bernardini, R.; Curatolo, P.; Galasso, C.; Pironi, V.; Coletta, M.; Marini, S. Effects of oral administration of common antioxidant supplements on the energy metabolism of red blood cells. Attenuation of oxidative stress-induced changes in Rett syndrome erythrocytes by CoQ10. Mol. Cell. Biochem. 2020, 463, 101–113. [CrossRef] 4. Wilson, M.C.; Trakarnsanga, K.; Heesom, K.J.; Cogan, N.; Green, C.; Toye, A.M.; Parsons, S.F.; Anstee, D.J.; Frayne, J. Comparison of the proteome of adult and cord erythroid cells, and changes in the proteome following reticulocyte maturation. Mol. Cell. Proteom. 2016, 15, 1938–1946. [CrossRef] 5. D’Alessandro, A.; Zolla, L. Biochemistry of red cell aging in vivo and storage lesions. Haematologica 2013, 7, 389–396. 6. van Wijk, R.; van Solinge, W.W. The energy-less red blood cell is lost: Erythrocyte enzyme abnormalities of glycolysis. Blood 2005, 106, 4034–4042. [CrossRef] Metabolites 2021, 11, 486 23 of 24

7. Yoo, J.-W.; Irvine, D.J.; Discher, D.E.; Mitragotri, S. Bio-inspired, bioengineered and biomimetic drug delivery carriers. Nat. Rev. Drug Discov. 2011, 10, 521–535. [CrossRef][PubMed] 8. Fang, R.H.; Hu, C.-M.J.; Zhang, L. Nanoparticles disguised as red blood cells to evade the immune system. Expert Opin. Biol. Ther. 2012, 12, 385–389. [CrossRef][PubMed] 9. Anselmo, A.C.; Mitragotri, S. Cell-mediated delivery of nanoparticles: Taking advantage of circulatory cells to target nanoparticles. J. Control Release 2014, 190, 531–541. [CrossRef][PubMed] 10. Tan, S.; Wu, T.; Zhang, D.; Zhang, Z. Cell or cell membrane-based drug delivery systems. Theranostics 2015, 5, 863. [CrossRef] 11. Han, X.; Wang, C.; Liu, Z. Red blood cells as smart delivery systems. Bioconjug. Chem. 2018, 29, 852–860. [CrossRef] 12. Goodacre, R. Metabolomics of a Superorganism 1–3. J. Nutr. Int. Res. Conf. Food 2007, 137, 259–266. [CrossRef] 13. Fiehn, O.; Kopka, J.; Dörmann, P.; Altmann, T.; Trethewey, R.N.; Willmitzer, L. Metabolite profiling for plant functional genomics. Nat. Biotechnol. 2000, 18, 1157–1161. [CrossRef][PubMed] 14. Beckonert, O.; Keun, H.C.; Ebbels, T.M.D.; Bundy, J.G.; Holmes, E.; Lindon, J.C.; Nicholson, J.K. Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts. Nat. Protoc. 2007, 2, 2692–2703. [CrossRef][PubMed] 15. Teng, R.; Junankar, P.R.; Bubb, W.A.; Rae, C.; Mercier, P.; Kirk, K. Metabolite profiling of the intraerythrocytic malaria parasite Plasmodium falciparum by 1H NMR spectroscopy. NMR Biomed. Int. J. Devoted Dev. Appl. Magn. Reson. In Vivo 2009, 22, 292–302. [CrossRef] 16. Pertinhez, T.A.; Casali, E.; Lindner, L.; Spisni, A.; Baricchi, R.; Berni, P. Biochemical assessment of red blood cells during storage by 1H nuclear magnetic resonance spectroscopy. Identification of a biomarker of their level of protection against oxidative stress. Blood Transfus. 2014, 12, 548–556. [CrossRef] 17. Nagana Gowda, G.A.; Gowda, Y.N.; Raftery, D. Expanding the limits of human blood metabolite quantitation using NMR spectroscopy. Anal. Chem. 2015, 87, 706–715. [CrossRef] 18. Kabeeruddin, M. Kitabul Advia; Aligarh Barqi Press: Delhi, India, 1937; Volume 2, p. 148. 19. Wasman, S.Q.; Mahmood, A.A.; Chua, L.S.; Alshawsh, M.A.; Hamdan, S. Antioxidant and gastroprotective activities of Andrographis paniculata (Hempedu Bumi) in Sprague Dawley rats. Indian J. Exp. Biol. 2011, 49, 767–772. 20. Yoopan, N.; Thisoda, P.; Rangkadilok, N.; Sahasitiwat, S.; Pholphana, N.; Ruchirawat, S.; Satayavivad, J. Cardiovascular effects of 14-Deoxy-11,12-didehydroandrographolide and Andrographis paniculata extracts. Planta Med. 2007, 73, 503–511. [CrossRef] [PubMed] 21. Sheeja, K.; Kuttan, G. Activation of Cytotoxic T Lymphocyte Responses and Attenuation of Tumor Growth in vivo by Andro- graphis paniculata Extract and Andrographolide. Immunopharmacol. Immunotoxicol. 2007, 29, 81–93. [CrossRef] 22. Banerjee, M.; Chattopadhyay, S.; Choudhuri, T.; Bera, R.; Kumar, S.; Chakraborty, B.; Mukherjee, S.K. Cytotoxicity and cell cycle arrest induced by andrographolide lead to programmed cell death of MDA-MB-231 breast cancer cell line. J. Biomed. Sci. 2016, 23, 40. [CrossRef] 23. Banerjee, M.; Parai, D.; Dhar, P.; Roy, M.; Barik, R.; Chattopadhyay, S.; Mukherjee, S.K. Andrographolide induces oxidative stress-dependent cell death in unicellular protozoan parasite Trypanosoma brucei. Acta Trop. 2017, 176, 58–67. [CrossRef] [PubMed] 24. Dua, V.K.; Ojha, V.P.; Roy, R.; Joshi, B.C.; Valecha, N.; Devi, C.U.; Sharma, V.P.; Subbarao, S.K. Anti-malarial activity of some xanthones isolated from the roots of Andrographis paniculata. J. Ethnopharmacol. 2004, 95, 247–251. [CrossRef][PubMed] 25. Hossain, M.S.; Urbi, Z.; Sule, A.; Rahman, K.M.H. Andrographis paniculata (Burm. f.) Wall. ex Nees: A review of ethnobotany, phytochemistry, and pharmacology. Sci. World J. 2014, 2014, 274905. [CrossRef][PubMed] 26. Misra, P.; Pal, N.L.; Guru, P.Y.; Katiyar, J.C.; Srivastava, V.; Tandon, J.S. Antimalarial activity of Andrographis paniculata (Kalmegh) against Plasmodium berghei NK 65 in Mastomys natalensis. Int. J. Pharmacogn. 1992, 30, 263–274. [CrossRef] 27. Niranjan, A.; Tewari, S.K.; Lehri, A. Biological activities of Kalmegh (Andrographis paniculata Nees) and its active principles-A review. Indian J. Nat. Prod. Resour. 2010, 1, 125–135. 28. Mishra, K.; Dash, A.P.; Dey, N. Andrographolide: A novel antimalarial diterpene lactone compound from Andrographis paniculata and its interaction with curcumin and artesunate. J. Trop. Med. 2011, 2011, 579518. [CrossRef][PubMed] 29. Zaridah, M.Z.; Idid, S.Z.; Wan Omar, A.; Khozirah, S. In vitro antifilarial effects of three plant species against adult worms of subperiodic Brugia malayi. J. Ethnopharmacol. 2001, 78, 79–84. [CrossRef] 30. Zaid, O.I.; Majid, R.A.; Sabariah, M.N.; Hasidah, M.S.; Al-Zihiry, K.; Yam, M.F.; Basir, R. Andrographolide effect on both Plasmodium falciparum infected and non infected RBCs membranes. Asian Pac. J. Trop. Med. 2015, 8, 507–512. [CrossRef] 31. Widyawaruyanti, A.; Astrianto, D.; Ilmi, H.; Tumewu, L.; Setyawan, D.; Widiastuti, E.; Dachliyati, L.; Tantular, I.S.; Hafid, A.F. Antimalarial activity and survival time of Andrographis paniculata fraction (AS202-01) on Plasmodium berghei infected mice. Res. J. Pharm. Biol. Chem. Sci. 2017, 8, 49–54. 32. Cossum, P.A. Role of the red blood cell in drug metabolism. Biopharm. Drug Dispos. 1988, 9, 321–336. [CrossRef] 33. Mohanty, J.; Nagababu, E.; Rifkind, J.M. Red blood cell oxidative stress impairs oxygen delivery and induces red blood cell aging. Front. Physiol. 2014, 5, 84. [CrossRef][PubMed] 34. Chaleckis, R.; Murakami, I.; Takada, J.; Kondoh, H.; Yanagida, M. Individual variability in human blood metabolites identifies age-related differences. Proc. Natl. Acad. Sci. USA 2016, 113, 4252–4259. [CrossRef] Metabolites 2021, 11, 486 24 of 24

35. Messana, I.; Ferroni, L.; Misiti, F.; Girelli, G.; Pupella, S.; Castagnola, M.; Zappacosta, B.; Giardina, B. Blood bank conditions and RBCs: The progressive loss of metabolic modulation. Transfusion 2000, 40, 353–360. [CrossRef][PubMed] 36. Yoshida, T.; Prudent, M.; D’alessandro, A. Red blood cell storage lesion: Causes and potential clinical consequences. Blood Transfus. 2019, 17, 27–52. [CrossRef] 37. Worley, B.; Powers, R. Multivariate Analysis in Metabolomics. Curr. Metab. 2013, 1, 92–107. 38. Eriksson, L.; Byrne, T.; Johansson, E.; Tyrgg, J.; Vikström, C. Multi-and Megavariate Data Analysis Basic Principles and Applications; Umetrics Academy, Sweden: Malmö, Sweden, 2006; Volume 1, ISBN 9197373052. 39. Aster, R.H. Adverse drug reactions affecting blood cells. Advers. Drug React. 2010, 196, 57–76. 40. Chong, J.; Wishart, D.S.; Xia, J. Using MetaboAnalyst 4.0 for Comprehensive and Integrative Metabolomics Data Analysis. Curr. Protoc. Bioinform. 2019, 68, e86. [CrossRef] 41. Kuhn, V.; Diederich, L.; Keller, S.T.C., IV; Kramer, C.M.; Lückstädt, W.; Panknin, C.; Suvorava, T.; Isakson, B.E.; Kelm, M.; Cortese-Krott, M.M. Red Blood Cell Function and Dysfunction: Redox Regulation, Nitric Oxide Metabolism, Anemia. Antioxid. Redox Signal. 2017, 26, 718–742. [CrossRef] 42. Bianconi, E.; Piovesan, A.; Facchin, F.; Beraudi, A.; Casadei, R.; Frabetti, F.; Vitale, L.; Pelleri, M.C.; Tassani, S.; Piva, F. An estimation of the number of cells in the human body. Ann. Hum. Biol. 2013, 40, 463–471. [CrossRef] 43. Kaur, H.; Clarke, S.; Lalani, M.; Phanouvong, S.; Guérin, P.; McLoughlin, A.; Wilson, B.K.; Deats, M.; Plançon, A.; Hopkins, H.; et al. Fake anti-malarials: Start with the facts. Malar. J. 2016, 15, 86. [CrossRef] 44. D’Alessandro, A.; Zolla, L. Proteomic analysis of red blood cells and the potential for the clinic: What have we learned so far? Expert Rev. Proteom. 2017, 14, 243–252. [CrossRef] 45. Bordbar, A.; Jamshidi, N.; Palsson, B.O. iAB-RBC-283: A proteomically derived knowledge-base of erythrocyte metabolism that can be used to simulate its physiological and patho-physiological states. BMC Syst. Biol. 2011, 5, 110. [CrossRef] 46. Divino Filho, J.C.; Barany, P.; Stehle, P.; Fürst, P.; Bergström, J. Free amino-acid levels simultaneously collected in plasma, muscle, and erythrocytes of uraemic patients. Nephrol. Dial. Transplant. 1997, 12, 2339–2348. [CrossRef] 47. Chesney, R.W.; Budreau, A.M. Chloroquine, a novel inhibitor of amino acid transport by rat renal brush border membrane vesicles. Amino Acids 1995, 8, 141–158. [CrossRef][PubMed] 48. Raftos, J.E.; Whillier, S.; Kuchel, P.W. Glutathione synthesis and turnover in the human erythrocyte: Alignment of a model based on detailed enzyme kinetics with experimental data. J. Biol. Chem. 2010, 285, 23557–23567. [CrossRef][PubMed] 49. Halliwell, B.; Gutteridge, J.M.C. Free Radicals in Biology and Medicine; Oxford University Press: New York, NY, USA, 2015; ISBN 0198717482. 50. Rabenstein, D.L. 1H NMR methods for the noninvasive study of metabolism and other processes involving small molecules in intact erythrocytes. J. Biochem. Biophys. Methods 1984, 9, 277–306. [CrossRef] 51. Reglinski, J.; Smith, W.E.; Brzeski, M.; Marabani, M.; Sturrock, R.D. Clinical analysis by proton spin-echo NMR. 2. Oxidation of intracellular glutathione as a consequence of penicillamine therapy in rheumatoid arthritis. J. Med. Chem. 1992, 35, 2134–2137. [CrossRef][PubMed] 52. Reglinski, J.; Smith, W.E.; Wilson, R.; Buchanan, L.M.; McKillop, J.H.; Thomson, J.A.; Brzeski, M.; Marabani, M.; Sturrock, R.D. Clinical analysis in intact erythrocytes using 1H spin echo NMR. Clin. Chim. Acta 1991, 201, 45–57. [CrossRef] 53. Rossi, R.; Milzani, A.; Dalle-Donne, I.; Giustarini, D.; Lusini, L.; Colombo, R.; Di Simplicio, P. Blood glutathione disulﬁde: In vivo factor or in vitro artifact? Clin. Chem. 2002, 48, 742–753. [CrossRef] 54. Krithika, R.; Verma, R.J.; Shrivastav, P.S. Antioxidative and cytoprotective effects of andrographolide against CCl4-induced hepatotoxicity in HepG2 cells. Hum. Exp. Toxicol. 2013, 32, 530–543. [CrossRef] 55. Pompella, A.; Visvikis, A.; Paolicchi, A.; De Tata, V.; Casini, A.F. The changing faces of glutathione, a cellular protagonist. Biochem. Pharmacol. 2003, 66, 1499–1503. [CrossRef] 56. Alberts, B.; Johnson, A.; Lewis, J.; Raff, M.; Roberts, K.; Walter, P. How Cells Obtain Energy from Food. In Molecular Biology of the Cell, 4th ed.; Garland Science: New York, NY, USA, 2002. Available online: https://www.ncbi.nlm.nih.gov/books/NBK26882 (accessed on 23 March 2021). 57. Schurr, A. Lactate, Not Pyruvate, Is the End Product of Glucose Metabolism via Glycolysis; Carbohydrate: Rijeka, Croatia, 2017; pp. 21–35. 58. Trager, W.; Jensen, J.B. Human malaria parasites in continuous culture. Sciences 1976, 193, 673–675. [CrossRef][PubMed] 59. Ashraf, A.; Ibraheem, Z.O.; Bello, R.O.; Unyah, N.Z.; Alhassan, A.S.; Nordin, N.; Ismail, I.S.; Wana, M.N.; Majid, R.A.; Basir, R. Evaluating the Time Dependent Anti-plasmodium Activity of Andrographolide and Chloroquine on Different Stages of the Intraerythrocytic Cycle of Plasmodium Falciparum 3D7 in Vitro. Res. Sq. 2020, 1–14. [CrossRef] 60. Geamanu, A.; Gupta, S.V.; Bauerfeld, C.; Samavati, L. Metabolomics connects aberrant bioenergetic, transmethylation, and gut microbiota in sarcoidosis. Metabolomics 2016, 12, 35. [CrossRef][PubMed] 61. Le Guennec, A.; Tayyari, F.; Edison, A.S. Alternatives to nuclear overhauser enhancement spectroscopy presat and carr–purcell– meiboom–gill presat for NMR-based metabolomics. Anal. Chem. 2017, 89, 8582–8588. [CrossRef] 62. Wishart, D.S.; Feunang, Y.D.; Marcu, A.; Guo, A.C.; Liang, K.; Vázquez-Fresno, R.; Sajed, T.; Johnson, D.; Li, C.; Karu, N. HMDB 4.0: The human metabolome database for 2018. Nucleic Acids Res. 2018, 46, D608–D617. [CrossRef] 63. An, L.; Shi, Q.; Feng, F. Metabolomics approach to identify therapeutically potential biomarkers of the Zhi-Zi-Da-Huang decoction effect on the hepatoprotective mechanism. RSC Adv. 2015, 5, 84048–84055. [CrossRef]