The Metabolic Impact of Two Different Parenteral Nutrition Lipid Emulsions in Children after Hematopoietic Stem Cell Transplantation: A Lipidomics Investigation

  Citation: Rangel-Huerta, O.D.; de la Torre-Aguilar, M.J.; Mesa, M.D.; Flores-Rojas, K.; Pérez-Navero, J.L.; Baena-Gómez, M.A.; Gil, A.; Gil-Campos, M. The Metabolic Impact of Two Different Parenteral Nutrition Lipid Emulsions in Children after Hematopoietic Stem Cell Transplantation: A Lipidomics Investigation. Int. J. Mol. Sci. 2022,23, 3667. https://doi.org/10.3390/ ijms23073667 Academic Editor: Alessandra Ferramosca Received: 28 January 2022 Accepted: 22 March 2022 Published: 27 March 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). International Journal of Molecular Sciences Article The Metabolic Impact of Two Different Parenteral Nutrition Lipid Emulsions in Children after Hematopoietic Stem Cell Transplantation: A Lipidomics Investigation Oscar Daniel Rangel-Huerta 1,† , María Joséde la Torre-Aguilar 2,† , María Dolores Mesa 3,4 , Katherine Flores-Rojas 2, Juan Luis Pérez-Navero 2, María Auxiliadora Baena-Gómez 2, Angel Gil 3,4,*,‡ and Mercedes Gil-Campos 2,5,‡ 1Section of Chemistry and Toxinology, Norwegian Veterinary Institute, P.O. Box 64, N-1431 Ås, Norway; oscar[email protected] 2Department of Pediatrics, Unit of Pediatric Research, Reina Sofia University Hospital, Maimonides Institute of Biomedical Research of Cordoba (IMIBIC), University of Córdoba, Avda Menéndez Pidal s/n, 14004 Cordoba, Spain; delatorr[email protected] (M.J.d.l.T.-A.); [email protected] (K.F.-R.); juanpereznaver[email protected] (J.L.P.-N.); [email protected] (M.A.B.-G.); [email protected] (M.G.-C.) 3 Department of Biochemistry and Molecular Biology II, Institute of Nutrition and Food Technology, Center of Biomedical Research, University of Granada, Avda. del Conocimiento s/n, 18016 Armilla, Spain; [email protected] 4Instituto de Investigación Biosanitaria ibs.Granada, 18012 Granada, Spain 5CIBEROBN (Physiopathology of Obesity and Nutrition), Institute of Health Carlos III (ISCIII), 28029 Madrid, Spain *Correspondence: [email protected] † These authors contributed equally to this work. ‡ These authors contributed equally to this work. Abstract: Hematopoietic stem cell transplantation (HSCT) involves the infusion of either bone marrow or blood cells preceded by toxic chemotherapy. However, there is little knowledge about the clinical benefits of parenteral nutrition (PN) in patients receiving high-dose chemotherapy during HSCT. We investigated the lipidomic profile of plasma and the targeted fatty acid profiles of plasma and erythrocytes in children after HSCT using PN with either a fish oil-based lipid emulsion or a classic soybean oil emulsion. An untargeted liquid chromatography high-resolution mass spectrometry platform connected with a novel in silico annotation algorithm was utilized to determine the most relevant chemical subclasses affected. In addition, we explored the interrelation between the lipidomics profile in plasma, the targeted fatty acid profile in plasma and erythrocytes, several biomarkers of inflammation, and antioxidant defense using an innovative data integration analysis based on Latent Components. We observed that the fish oil-based lipid emulsion had an impact in several lipid subclasses, mainly glycerophosphocholines (PC), glycerophosphoserines (PS), glycerophosphoethanolamines (PE), oxidized PE (O-PE), 1-alkyl,2-acyl PS, lysophosphatidylethanolamines (LPE), oxidized PS (O-PS) and dicarboxylic acids. In contrast, the classic soybean oil emulsion did not. Several connections across the different blocks of data were found and aid in interpreting the impact of the lipid emulsions on metabolic health. Keywords: bone marrow transplantation; fat emulsions; intravenous; fatty acids; omega-3; hematopoietic stem cell transplantation; lipidomics; parenteral nutrition 1. Introduction Hematopoietic stem cell transplantation (HSCT) is currently the standard of care for many malignant and nonmalignant blood diseases. HSCT involves the infusion of either bone marrow, peripheral blood or cord blood as a stem cell source preceded by toxic Int. J. Mol. Sci. 2022,23, 3667. https://doi.org/10.3390/ijms23073667 https://www.mdpi.com/journal/ijms Int. J. Mol. Sci. 2022,23, 3667 2 of 20 chemotherapy. The conditioning regimen decreases the tumor burden and maximizes the donor cells’ capability to engraft successfully by suppressing the patient’s immune system [ 1 ]. This aggressive chemotherapy affects the digestive system and limits oral food intake. Thus, a majority of patients undergoing HSCT suffer from severe mucositis and enteritis due to cytotoxic therapy and immune dysregulation, resulting in prolonged decreased oral intake, nausea, vomiting, and diarrhea. In this regard, nutrition support is often required during HSCT, given the gastrointestinal toxicity that frequently precludes adequate protein-calorie intake [ 2 ]. Parenteral nutrition (PN) is often given to patients to maintain their nutritional status during the peritransplant period, and it is reserved for those patients who are unable to tolerate enteral feedings [ 3 , 4 ]. However, the clinical benefits of PN in patients receiving high-dose chemotherapy during HSCT are unknown [ 5 , 6 ]. An integrative review on the efficacy of enteral nutrition and PN for meeting the nutrition and energy needs of pediatric patients following HSCT was published in 2019 [ 7 ]. More recently, the Pediatric Diseases Working Party (PDWP) of the European Society for Blood and Marrow Transplantation (EBMT) reported the resulting consensus on the management of sinusoidal obstructive syndrome, mucositis, enteral and parenteral nutrition, iron overload, and emesis during HSCT [ 8 ], as well as on prevention of infections [ 9 ]. Moreover, early and customized nutritional intervention may be optimal for all patients who undergo HSCT to ameliorate body weight loss associated with nutrition-related adverse events [ 10 ]. Previous studies have demonstrated that the use of lipid-based PN after allogeneic bone marrow transplantation is associated with a lower incidence of lethal acute graftversus-host-disease (aGvHD) and hyperglycemia, without negatively affecting the time to engraftment of infused cells [ 11 , 12 ]. Those studies suggested that the intravenously administered lipids might have influenced the severity of disease by modulation of immune response and synthesis of eicosanoids that participate in the pathogenesis of aGvHD. Traditionally, the lipids used in artificial nutrition are based on vegetable oils, such as soybean oil (SO), which provide the essential fatty acids (FA) linoleic (LA, 18:2 ω -6) and α -linolenic acid (LNA, 18:3 ω -3) in relatively high amounts. These emulsions have raised concerns because of their potential adverse effects involving oxidative stress, inflammation, and immune response, probably due to the excess of unsaturated FAs [ 13 ]. In recent years, fish oil (FO) in lipid emulsions has been introduced as a component of PN. These emulsions contain eicosapentaenoic acid (EPA, 20:5 ω -3) and docosahexaenoic acid (DHA, 22:6 ω -3), which modulate the synthesis of eicosanoids, the activity of nuclear receptors and nuclear transcription factors, and the production of resolvins, protectins, and maresins, with recognized anti-inflammatory and immunomodulatory effects in critically ill patients [ 14 ]. Therefore, they have been associated with less hepatic toxicity and lower levels of lowdensity lipoprotein triacylglycerols and C-reactive protein compared to soybean lipid emulsions in postoperative patients [ 15 ]. They are likely to reduce infections, the length of hospital stays, and liver dysfunction without influencing mortality and may be a safe and preferable choice in post-surgery patients [16]. Several studies have revised the effects of PN in children with HSCT [ 3 , 4 , 7 ]. However, only a few addressed the question of the potential effects of enriched FO lipid emulsions compared with standard lipid emulsions [ 17 – 19 ]. We have previously shown that PN containing SO or long-chain polyunsaturated (LC-PUFA) ω -3 FAs enriched emulsions for ten days is safe for children [ 17 ]. FO-containing emulsion in long-term PN increases the levels of plasma ω -3 LC-PUFA and decreases those of arachidonic acid (20:4 ω -6) [ 17 ] and can improve the antioxidant profile by increasing levels of α -tocopherol in children who have a high risk of suffering oxidative stress and metabolic disorders [ 18 ]. In addition, previous findings of our group suggest that different lipid emulsions in PN administered to children undergoing HSCT for a short period do not induce significant inflammatory changes. However, ω -3 LC-PUFA-supplemented PN for more than 21 days may modulate the inflammatory response [19]. Lipids are essential metabolites engaged in several cellular functions; thus, they directly monitor cellular metabolic status. Within the omics sciences context, the complete Int. J. Mol. Sci. 2022,23, 3667 3 of 20 lipid profile and content in a cell is called lipidome, and the study itself lipidomics. The lipids classification comprises classes and subclasses depending on the head group and the linkage type between the head group and aliphatic chains [ 20 ]. A comprehensive work reviewing their functionality can be consulted elsewhere [ 21 ]. Some studies carried out in adults with HSCT have shown an altered metabolic profile caused both by the disease and its immunosuppressive treatment [ 22 ], and several plasma biomarkers have been claimed to help predict the risk of aGvHD in clinical settings [ 23 – 26 ]. However, no metabolomic nor lipidomics approaches have been reported in children with HSCT. Therefore, the present study aimed to investigate the plasmatic lipidomics profile in children after HSCT using an FO lipid emulsion and compare the results with those for the classic SO emulsion used in PN. In addition, we investigated the interrelation between the lipidomics profile in plasma, the targeted FA profile in plasma and erythrocytes, several biomarkers of inflammation, and antioxidant defense. The investigated children with HSCT were the same that served for previous publications [17–19]. 2. Results 2.1. General Characteristics of the Patients Table 1presents a summary of demographics and clinical complications in the population here included. All the patients presented mucositis without significant differences in the degrees of severity (p= 0.121). The days of fever post HSTC transplantation were similar in both groups (p= 0.204). No patient presented renal pathology or development of hepatic veno-occlusive disease (VOD). Excluding the two patients who received an autologous transplant and therefore cannot be affected by aGvHD, three patients in the SO-based parenteral lipid emulsion (SOPLE) developed GvHD (75%) versus two patients (50%) in the FO-based parenteral lipid emulsion (FOPLE) group (p= 0.467). No patient died during admission, and there was a 100-day survival rate of 100% in both groups. PN was established to prevent or correct the adverse effects of malnutrition and was generally started in these patients when the intake was less than 2/3 of the basal energy needs. It could not be administered enterally, and its need was anticipated by a period greater than 5 to 7 days. In our sample, it began within a range of 1 to 3 days after BMT infusion: PN was administered through a central venous catheter, which all these patients had implanted prior to BMT. Withdrawal of PN was performed when the pathology that caused its onset improved, generally mucositis, and when oral intake reaches 2/3 of the estimated nutritional requirements. Table 1. Demographic characteristics and clinical complications in 10 children undergoing hematopoietic stem cell transplantation (HSCT) with two different lipid emulsions in parenteral nutrition, soybean oil, and n-3 PUFA-groups. SOPLE FOPLE Sex (male/female) 2/2 2/4 Age (months) 101.5 (8–180) 90.5 (31–132) Pathology Hematologic diseases 4 4 Solid tumors 2 Type of HSCT Allogeneic 4 4 Autologous 2 GVHD 3 2 VOD 0 0 Time of engraftment PMN: PMN > 500/mm315.5 (14–21) 13 (11–20) Platelets > 20,000 17.5 (15–24) 15 (12–68) Total days of PN 13 (11–25) 16 (9–24) Days of hospitalization 34 (31–37) 31 (29–43) HSTC: hematopoietic stem cell transplantation; GvHD: graft-versus-host disease; PMN: Polymorphonuclear cell count; VOD: veno-occlusive disease. Data are expressed as median and the minimum and maximum range. Int. J. Mol. Sci. 2022,23, 3667 4 of 20 2.2. Lipidomics Results Only samples from the entire cohort were included from six subjects following PN with SOPLE and six with FOPLE for the lipidomics analysis. After running the lipidomics and preprocessing the obtained raw data, we detected 612 features in positive mode and 1038 in negative mode. The feature-clustering function identified 235 clusters of two or more features. Thus, the final analysis included 1005 variables corresponding to single features or clusters of features. A principal component analysis (PCA) was used for an initial exploration to assess the quality of the lipidomics data and to detect the presence of outliers (Figure 1). The PCA revealed no outliers, and thus, all the samples were included for further analysis. Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW 4 of 21 Days of hospitalization 34 (31–37) 31 (29–43) HSTC: hematopoietic stem cell transplantation; GvHD: graft-versus-host disease; PMN: Polymorphonuclear cell count; VOD: veno-occlusive disease. Data are expressed as median and the minimum and maximum range. 2.2. Lipidomics Results Only samples from the entire cohort were included from six subjects following PN with SOPLE and six with FOPLE for the lipidomics analysis. After running the lipidomics and preprocessing the obtained raw data, we detected 612 features in positive mode and 1038 in negative mode. The feature-clustering function identified 235 clusters of two or more features. Thus, the final analysis included 1005 variables corresponding to single features or clusters of features. A principal component analysis (PCA) was used for an initial exploration to assess the quality of the lipidomics data and to detect the presence of outliers (Figure 1). The PCA revealed no outliers, and thus, all the samples were included for further analysis. Figure 1. PCA scores plot of the model, including samples from SO-based parenteral lipid emulsion (SOPLE) and FO-based (FOPLE) parenteral lipid emulsion groups at baseline. 2.2.1. Differences at Baseline in the Lipidomics Profile An orthogonal partial least squares discriminant analysis (OPLS-DA) was implemented to detect if the groups showed a significant difference at baseline. The OPLS-DA is a multivariate supervised model that discriminates between groups, as it tries to maximize the difference among classes using the class as a reference. In this case, the model comparing both groups at baseline was of poor quality, as shown by the low Q2 and R2 (0.423 and 0.326, respectively) and a not significant CV-ANOVA. Thus, we discarded systematic differences between groups before the intervention. Additionally, we ran a t-test among groups, observing no significant differences in any detected metabolites at baseline. 2.2.2. Differences after the Intervention in the Lipidomics Profile The next step was to detect any differences between the groups associated with the intervention. We utilized an analysis of variance multiblock partial least squares (AMOPLS) for such a purpose. Table 2 summarized the AMOPLS output and revealed that treatment and treatment x time interaction’s main effects were significant as represented by Figure 1. PCA scores plot of the model, including samples from SO-based parenteral lipid emulsion (SOPLE) and FO-based (FOPLE) parenteral lipid emulsion groups at baseline. 2.2.1. Differences at Baseline in the Lipidomics Profile An orthogonal partial least squares discriminant analysis (OPLS-DA) was implemented to detect if the groups showed a significant difference at baseline. The OPLS-DA is a multivariate supervised model that discriminates between groups, as it tries to maximize the difference among classes using the class as a reference. In this case, the model comparing both groups at baseline was of poor quality, as shown by the low Q 2 and R 2 (0.423 and 0.326, respectively) and a not significant CV-ANOVA. Thus, we discarded systematic differences between groups before the intervention. Additionally, we ran a t-test among groups, observing no significant differences in any detected metabolites at baseline. 2.2.2. Differences after the Intervention in the Lipidomics Profile The next step was to detect any differences between the groups associated with the intervention. We utilized an analysis of variance multiblock partial least squares (AMOPLS) for such a purpose. Table 2summarized the AMOPLS output and revealed that treatment and treatment x time interaction’s main effects were significant as represented by the RSR and R2Yp-values (p≤0.05). It is relevant to mention that the residuals accounted for 75% of the total observed variability. This is a high percentage and shows that other analytical or biological factors might be more relevant in the variation than those here studied. According to those results, it is necessary to analyze each group’s data independently. Then, two independent OPLS-DA models were built to investigate the differences between the initial and the final samples on each group. On the one hand, it was not possible to generate a significant model for the FOPLE group. On the other hand, the model Int. J. Mol. Sci. 2022,23, 3667 5 of 20 corresponding to the SOPLE group appeared to be significant (p= 0.04). However, when the permutation test plot (Supplementary Figure S1) was inspected, it revealed that the model might be spurious. This can be confirmed by the relatively high difference between the R2Y and the Q2(0.947 and 0.717, respectively). Table 2. AMOPLS output from the comparison between SO-based parenteral lipid emulsion (SOPLE) and FO-based parenteral lipid emulsion (FOPLE) before and after the intervention using treatment, time, and interaction as main effects using the lipidomics data. Effect Name RSS RSR RSR p-Value R2Y p-Value Tp1 Tp2 Tp3 To1 Treatment 0.08 1.159 0.03 0.03 0.046 0.042 0.825 0.239 Time 0.081 1.098 0.81 0.01 0.049 0.869 0.058 0.252 Treatment ×Time 0.085 1.196 0.05 0.01 0.851 0.041 0.053 0.232 Residuals 0.753 1 NA NA 0.054 0.049 0.064 0.277 NA: Not calculated; RSS: Relative Sum of Squares; RSR: Residual Structure Ratio; To: Orthogonal Component; Tp: Predictive Component. Our next step was to run a univariate analysis, specifically, a t-test comparing the basal samples versus the final samples for each group independently. Then, the output was used to feed the mummichog algorithm [ 27 ] included in the MS-Peaks to pathways function embedded in MetaboAnalyst [ 28 ]. This function takes advantage of the highresolution mass spectrometry data generated and provides some guidance for biological interpretation when fragmentation data is unavailable. Such an analysis revealed that the most relevant classes affected in the FOPLE group were glycerophosphocholines (PC), glycerophosphoserines (PS), oxidized glycerophosphoethanolamines (O-PE), glycerophosphoethanolamines (PE), 1-alkyl,2-acyl-PS, lysophosphatidylethanolamines (LPE), oxidized PS (O-PS), and dicarboxylic acids that showed an upward trend after the intervention (all p ≤ 0.05). In contrast, oxidized glycerophosphocholines (O-PC), phosphocholine ceramides (PC-Cer) or sphingomyelins (SM), phosphoethanolamine ceramides (PE-Cer), and ceramide-1-Phosphate (Cer-1-P) presented a downward trend after the intervention (all p ≤ 0.05). Besides, the algorithm could not provide relevant classes for the SOPLE group, possibly due to the low number of significant features. A detailed list of the statistical output from the algorithm and the compounds assignment is attached as Supplemental Material. 2.3. Differences after the Intervention in the General Biochemistry, Inflammatory Biomarkers, and FAs Profiles in Erythrocytes Descriptive data from the general biochemistry, inflammatory biomarkers, and FA profiles in erythrocytes is presented in Tables 3and 4. The figures revealed that ApoA and HDL decreased after the SOPLE intervention whereas ApoB increased. While, in the FOPLE group, a significant increase in GGT and LDL in plasma (Table 2) and EPA, total PUFA ω-3 and Index ω-3 in erythrocytes were observed. As we have a low nand many variables, we analyzed using the AMOPLS approach. For such analysis, the variables analyzed in erythrocytes were treated as one block. We observed no significant main effects in samples from erythrocytes. Nevertheless, EPA and total ω -3 PUFA increased notably after the treatment in the FOPLE group, compared with the SOPLE group (p< 0.05). However, the results from plasma revealed a significant time effect (RSR p-value = 0.01) but not a treatment or interaction effect. Figure 2shows the top 10 variables that discriminate samples ordered according to the variable of influence in the projection (VIP 2 ) value associated with the time main effect. It is worth noting that one of the advantages of using AMOPLS is that we can obtain the influence of each variable on each factor, e.g., we observed that variables, such as arachidonic acid, tumor necrosis factor (TNFα ), and alanine-aminotransferase (ALT), have a high VIP 2 in the time factor but also present a relevant influence on the treatment x time interaction, though this was not significant but need to be considered. Int. J. Mol. Sci. 2022,23, 3667 6 of 20 Table 3. General biochemistry and inflammatory biomarker values in plasma from subjects in SObased parenteral lipid emulsion (SOPLE) and FO-based parenteral lipid emulsion (FOPLE) formula groups before and after the intervention. SOPLE (n= 4) FOPLE (n= 6) Basal Final Basal Final Median Min Max Median Min Max Median Min Max Median Min Max Glucose (mg/dL) 90.5 63.0 103.0 94.0 77.0 208.0 94.5 68.0 103.0 91.5 82 120 Urea (mg/dL) 19.0 13.0 27.0 26.0 24.0 55.0 17.0 10.0 21.0 30.0 12 49 Creatinine (mg/dL) 0.40 0.30 0.53 0.43 0.35 0.68 0.48 0.35 0.55 0.42 0.33 0.52 AST (U/L) 44.0 19.0 114.0 41.5 15.0 81.0 36.0 17.0 82.0 32.5 25 71 ALT (U/L) 44.5 14.0 228.0 57.0 18.0 99.0 45.5 10.0 127.0 34.0 16 59 GGT (U/L) 19.0 9.0 46.0 36.0 19.0 161.0 28.5 17.0 77.0 102.0 a59 267 ALP (U/L) 105.5 87.0 220.0 144.5 117.0 332.0 155.5 102.0 182.0 160.5 106 332 ApoA (mg/dL) 98.5 71.0 105.0 56.0 a50.0 77.0 98.5 60.0 124.0 65.0 54 77 ApoB (mg/dL) 73.0 42.0 89.0 127.0 a53.0 145.0 97.5 54.0 205.0 120.5 88 222 Total cholesterol (mg/dL) 145 112 173 214 97 232 189 123 292 209 176 366 HDL (mg/dL) 35.0 26.0 42.0 15.0 a12.0 23.0 32.0 15.0 45.0 18.5 13 24 LDL (mg/dL) 13.5 9.0 19.0 110.5 49.0 181.0 15.5 0.0 38.0 131.0 a109 163 Triacylglycerols (mg/dL) 135 79 176 308 153 418 107 91 445 299 161 588 Bilirubin (mg/dL) 0.80 0.30 1.40 0.70 0.10 1.90 0.45 0.30 0.90 1.00 a0.5 1.7 Data presented correspond only to those patients that had the entire data blocks. a Significantly different in a t-test comparing basal vs. final time considering a p-value < 0.05 as the cut-off. ALT: alanine-aminotransferase; ALP: alkaline phosphatase; Apo: apolipoprotein; AST: aspartate aminotransferase; HDL: high-density cholesterol; GGT: gamma-glutamyl transferase; LDL: low-density cholesterol; FOPLE: FO-based formula; PCR: protein C-reactive; SOPLE: SO-based formula. Table 4. Fatty acid profile of the red blood cell membrane from subjects in SO-based parenteral lipid emulsion (SOPLE) and FO-based parenteral lipid emulsion (FOPLE) formula groups before and after the intervention. SOPLE (n= 4) FOPLE (n= 6) Basal Final Basal Final Fatty Acids, % Relative to Total FAs 1Median Min Max Median Min Max Median Min Max Median Min Max Myristic acid (C14:0) 0.80 0.44 1.76 0.66 0.49 1.10 0.53 0.36 3.03 0.63 0.37 0.96 Palmitic acid (C16:0) 24.70 23.60 26.10 24.10 22.10 25.0 23.30 22.70 28.30 23.75 22.60 25.50 Palmitoleic acid (C16:1) 0.49 0.37 0.70 0.19 0.00 0.63 0.39 0.00 1.18 0.25 0.00 0.83 Margaric acid (C17:0) 0.39 0.33 0.60 0.87 a0.45 0.98 0.39 0.00 1.27 0.16 0.00 0.90 Estearic acid (C18:0) 15.85 15.70 17.00 15.80 15.30 20.3 15.95 13.00 17.30 15.70 14.90 17.70 Oleic acid (C18:1n-9c) 15.35 13.00 20.50 15.30 14.40 16.2 14.85 13.70 25.20 15.15 14.60 17.10 Vaccenic acid (C18:1n-7) 0.99 0.50 1.17 1.09 0.92 1.25 1.11 0.95 1.49 1.11 1.01 1.21 Linoleic acid (C18:2ω-6) 7.90 7.70 8.50 8.85 7.70 10.5 9.30 5.40 9.50 8.15 7.30 9.90 Arachidic acid (C20:0) 0.42 0.00 0.48 0.23 0.00 0.50 0.00 0.00 0.53 0.43 0.00 0.50 Linolenic acid (C18:3 ω -3) 0.16 0.00 0.41 0.00 0.00 0.33 0.00 0.00 0.40 0.00 0.00 0.00 Behenic acid (C22:0) 1.63 1.21 1.86 1.59 1.54 1.72 1.90 1.17 3.97 1.72 1.55 1.94 Dihomo-γ-linolenic acid (C20:3ω-6) 1.78 1.69 1.95 1.69 1.51 1.93 1.41 0.00 2.39 1.33 1.16 2.21 Dihomo-α-linolenic (C20:3ω-3) 0.61 0.57 0.73 0.75 0.62 1.52 0.54 0.42 0.85 0.66 0.38 1.10 Arachidonic acid (C20:4ω-6) 14.80 10.30 17.60 14.45 13.10 15.8 14.65 9.70 16.40 13.65 12.80 15.10 Eicosapentanoic acid (C20:5ω-3) 0.00 0.00 0.34 0.00 0.00 0.57 0.35 0.00 0.46 1.51 a0.46 2.12 Lingnoceric (24:0) 4.65 3.63 5.07 4.44 4.36 4.89 4.59 2.88 5.25 4.50 4.15 4.70 Nervonic acid (24:1n9) 3.76 3.06 5.14 3.69 2.93 4.47 3.47 2.08 6.11 3.25 2.95 3.88 Docosapentanoic acid (C22:5ω-3) 1.33 1.28 1.78 1.43 1.24 1.53 1.48 0.92 1.94 1.84 1.38 2.41 Int. J. Mol. Sci. 2022,23, 3667 7 of 20 Table 4. Cont. SOPLE (n= 4) FOPLE (n= 6) Basal Final Basal Final Fatty Acids, % Relative to Total FAs 1Median Min Max Median Min Max Median Min Max Median Min Max Docosahexaenoic acid (C22:6ω-3) 3.56 2.44 4.10 3.73 2.37 4.04 4.05 2.74 4.93 4.43 4.07 6.15 SFA 48.80 47.20 49.90 48.55 47.10 49.9 47.90 46.30 49.00 47.05 46.10 48.50 UFA 51.20 50.10 52.80 51.45 50.10 52.90 52.10 51.00 53.70 52.95 51.50 53.90 MUFA 20.65 19.50 24.70 20.15 19.40 21.60 19.80 18.70 29.90 19.65 19.50 22.10 SFA/MUFA ratio 2.35 2.00 2.50 2.45 2.20 2.50 2.40 1.60 2.60 2.40 2.20 2.50 DUFA 7.90 7.70 8.50 8.85 7.70 10.5 9.30 5.40 9.50 8.15 7.30 9.90 MUFA/DUFA ratio 2.65 2.50 2.90 2.30 1.90 2.8 2.30 2.00 5.60 2.40 2.10 2.80 PUFA 30.50 26.60 32.10 31.20 30.30 31.9 32.40 21.40 32.80 32.70 30.20 34.00 MUFA/PUFA ratio 0.65 0.60 0.90 0.70 0.60 0.70 0.60 0.60 1.40 0.60 0.60 0.70 PUFA ω-6 24.40 20.80 27.10 25.20 24.50 25.6 25.35 16.00 25.90 23.45 22.90 25.80 PUFA ω-3 5.80 4.40 6.90 6.15 5.10 6.60 6.65 4.70 7.80 8.70 a7.10 10.70 PUFA ω-6 >18 C 16.55 12.30 19.40 16.25 14.70 17.5 16.35 10.70 17.80 15.35 14.00 16.60 PUFA ω-3 >18 C 5.45 4.40 6.90 5.95 5.10 6.60 6.55 4.70 7.80 8.70 a7.10 10.70 UI 2.65 2.40 3.00 2.70 2.60 2.90 2.85 2.30 3.10 3.00 2.80 3.30 Ratio ω-6/ω-3 3.84 3.62 6.13 4.14 3.70 4.91 3.58 2.99 5.49 2.68 2.19 3.35 Index ω-3 3.56 2.44 4.44 3.87 2.37 4.34 4.44 2.74 5.36 6.14 a4.67 7.70 Delta9 desaturase 0.97 0.76 1.30 0.99 0.71 1.02 0.92 0.82 1.94 0.99 0.83 1.11 Delta6 desaturase 0.23 0.22 0.23 0.18 0.17 0.25 0.18 0.00 0.26 0.17 0.12 0.27 Delta5 desaturase 0.12 0.10 0.19 0.12 0.10 0.13 0.10 0.00 0.17 0.10 0.08 0.17 Data presented correspond only to those patients that had the entire data blocks. 1 The amount of each fatty acid was calculated as a percentage of the total fatty acid content (relative%). a Significantly different in a t-test comparing basal vs. final time considering a p-value of < 0.05 as the cut-off. DUFA: double unsaturated fatty acid; FOPLE: FO-based formula; MUFA: monounsaturated fatty acid; PUFA: polyunsaturated fatty acid; SFA: saturated fatty acid; SOPLE: SO-based formula; UFA: unsaturated fatty acid, UI: unsaturation index. Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW 8 of 21 Figure 2. Most discriminant features according to the amOPLS approach for the integrated analysis of the general biochemistry, FAs, and inflammatory biomarkers in plasma. The variables are ordered by the VIP2 value corresponding to the time effect. VIP2: variable of importance in the projection. The higher it is, the more influence on the model. 2.4. Data Integration Analysis for Biomarker Discovery Using Latent Components (DIABLO) A Data Integration Analysis for Biomarker Discovery using Latent Components (DIABLO) was conducted in an exploratory manner due to the low number of samples. Data were divided into four blocks: lipidomics in plasma, plasmatic measurements (including FAs and inflammatory biomarkers), FAs in erythrocytes, and antioxidant enzymes and vitamins (data previously published in Baena et al.). Unfortunately, two subjects from the SOPLE group were dropped due to the lack of data corresponding to the antioxidant enzymes at one of the time points. The outcome was to identify the association between the different blocks of data and the difference between both groups, SOPLE and FOPLE. Samples from basal and final points from each group were merged. DIABLO relies on identifying a limited number of correlated variables from multiple datasets to predict an outcome. In our study, the outcome was the «treatment». In brief, the method is an extension of sparse generalized canonical correlation analysis [29], which is a generalization of partial least squares for multiple matching datasets (Q) to a supervised learning framework [30,31]. The design matrix is a Q × Q matrix representing if and by how much each dataset should be correlated for the model’s algorithms in the DIABLO analysis. Values range from 0 to 1. For the present study, a value of 0.1 was utilized to pursue a maximum separation. Once the design matrix was assigned, a DIABLO model with two components was first fitted without any variable selection, and global performance was assessed using a 5-Mfold cross-validation. The model was tuned using internal functions of the mixOmics [31] package to find the optimal number of components and the optimal number of variables for each dataset. The main output of DIABLO is a set of components chosen in the model, a set of loading vectors, and a list of selected variables from each block of data associated with each component. Loadings are the coefficients assigned to each variable to define each component, and their absolute value represents the importance of each variable in DIABLO [30]. Figure 2. Most discriminant features according to the amOPLS approach for the integrated analysis of the general biochemistry, FAs, and inflammatory biomarkers in plasma. The variables are ordered by the VIP 2 value corresponding to the time effect. VIP 2 : variable of importance in the projection. The higher it is, the more influence on the model. Int. J. Mol. Sci. 2022,23, 3667 8 of 20 2.4. Data Integration Analysis for Biomarker Discovery Using Latent Components (DIABLO) A Data Integration Analysis for Biomarker Discovery using Latent Components (DIABLO) was conducted in an exploratory manner due to the low number of samples. Data were divided into four blocks: lipidomics in plasma, plasmatic measurements (including FAs and inflammatory biomarkers), FAs in erythrocytes, and antioxidant enzymes and vitamins (data previously published in Baena et al.). Unfortunately, two subjects from the SOPLE group were dropped due to the lack of data corresponding to the antioxidant enzymes at one of the time points. The outcome was to identify the association between the different blocks of data and the difference between both groups, SOPLE and FOPLE. Samples from basal and final points from each group were merged. DIABLO relies on identifying a limited number of correlated variables from multiple datasets to predict an outcome. In our study, the outcome was the «treatment». In brief, the method is an extension of sparse generalized canonical correlation analysis [ 29 ], which is a generalization of partial least squares for multiple matching datasets (Q) to a supervised learning framework [ 30 , 31 ]. The design matrix is a Q × Qmatrix representing if and by how much each dataset should be correlated for the model’s algorithms in the DIABLO analysis. Values range from 0 to 1. For the present study, a value of 0.1 was utilized to pursue a maximum separation. Once the design matrix was assigned, a DIABLO model with two components was first fitted without any variable selection, and global performance was assessed using a 5-Mfold cross-validation. The model was tuned using internal functions of the mixOmics [ 31 ] package to find the optimal number of components and the optimal number of variables for each dataset. The main output of DIABLO is a set of components chosen in the model, a set of loading vectors, and a list of selected variables from each block of data associated with each component. Loadings are the coefficients assigned to each variable to define each component, and their absolute value represents the importance of each variable in DIABLO [30]. The final optimized model included one component. Figure 3presents a correlation matrix across the different data blocks for the first dimension. It is possible to observe the strong correlation, according to the high R 2 , between the lipidomics dataset and the block corresponding to the plasmatic FAs and general biochemistry parameters. The latter also presented a strong correlation with the block corresponding to the erythrocyte FAs. The correlation between the lipidomics and the FAs in erythrocytes is strong (r 2 = 0.84). Finally, the association between the antioxidant enzymes and vitamins block appears less connected with the other data blocks. Int. J. Mol. Sci. 2022,23, 3667 9 of 20 Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW 9 of 21 The final optimized model included one component. Figure 3 presents a correlation matrix across the different data blocks for the first dimension. It is possible to observe the strong correlation, according to the high R2, between the lipidomics dataset and the block corresponding to the plasmatic FAs and general biochemistry parameters. The latter also presented a strong correlation with the block corresponding to the erythrocyte FAs. The correlation between the lipidomics and the FAs in erythrocytes is strong (r2 = 0.84). Finally, the association between the antioxidant enzymes and vitamins block appears less connected with the other data blocks. Figure 3. Correlation matrix corresponding to the unique component of the DIABLO model comparing the four data blocks. AOX: antioxidant enzyme activities, FA: fatty acids; plasma, measurements in plasma (including general biochemistry and FA). FOPLE: Fish-oil based formula; SOPLE: SO-based formula. The values presented correspond to the R2; a value closer to 1 reflects a better association between pairs. In addition, Figure 4A corresponds to a Circos diagram built on a similarity matrix [32] and represents the correlation between variables from different datasets. Pairs of variables presenting values above the cut-off (r = 0.4) are connected with red and blue inner lines according to the type of correlation (e.g., positive, or negative). We can observe that the main correlations include variables from all the data blocks. Remarkably, TNF-α is negatively correlated with EPA in plasma and EPA, the ω-3 Index, and total PUFA ω-3 in erythrocytes. Furthermore, we can observe a positive correlation between glutathione reductase and the ω-3 Index and EPA in plasma. The lipidomics block features are mostly correlated among them and connected with the other blocks, revealing that they might be FA-related metabolites as per their molecular weight. Besides, Figure 4B represents the loading weights of the most relevant variables for each data block according to the optimized model. The component loading values highlight induced variables concerning the experimental factor treatment. Figure 3. Correlation matrix corresponding to the unique component of the DIABLO model comparing the four data blocks. AOX: antioxidant enzyme activities, FA: fatty acids; plasma, measurements in plasma (including general biochemistry and FA). FOPLE: Fish-oil based formula; SOPLE: SO-based formula. The values presented correspond to the R 2 ; a value closer to 1 reflects a better association between pairs. In addition, Figure 4A corresponds to a Circos diagram built on a similarity matrix [ 32 ] and represents the correlation between variables from different datasets. Pairs of variables presenting values above the cut-off (r= 0.4) are connected with red and blue inner lines according to the type of correlation (e.g., positive, or negative). We can observe that the main correlations include variables from all the data blocks. Remarkably, TNFα is negatively correlated with EPA in plasma and EPA, the ω -3 Index, and total PUFA ω -3 in erythrocytes. Furthermore, we can observe a positive correlation between glutathione reductase and the ω -3 Index and EPA in plasma. The lipidomics block features are mostly correlated among them and connected with the other blocks, revealing that they might be FA-related metabolites as per their molecular weight. Besides, Figure 4B represents the loading weights of the most relevant variables for each data block according to the optimized model. The component loading values highlight induced variables concerning the experimental factor treatment. Int. J. Mol. Sci. 2022,23, 3667 16 of 20 injected randomly to run RP-UPLC-FTMS to acquire two LC-MS datasets with positive-ion and negative-ion detection, respectively, in two respective rounds of the sample injections. 4.6.2. Data Acquisition For all UPLC-FTMS, a Dionex Ultimate 3000 UHPLC system coupled to a Thermo Scientific LTQ-Orbitrap Velos Pro mass spectrometer equipped with an electrospray ionization source was used. RP-UPLC-FTMS runs were carried out for lipid analyses using a Waters BEH C8 UPLC column (2.1 × 50 mm, 1.7 µ m) for chromatographic separation. The mobile phase was (A) 0.01% formic acid in water and (B) 0.01% formic acid in acetonitrile-isopropanol (1:1, v/v). The efficient gradient was 20% to 55% B in 5 min, 55% to 100% B in 12.5 min, and 100% B for 2.5 min before the column was equilibrated for 3.5 min at 20% B between injections. The column flow rate was 400 µ L/min, and the column temperature was maintained at 60 ◦ C. For relative quantitation, the MS instrument was operated in the survey-scan mode with full-mass FTMS detection at a mass resolution of 60,000 full-width at half maximum (FWHM) @ mass to charge ratio (m/z) 200. The mass scan range was m/z80 to 1800 for both positive-ion and negative-ion detection. 4.6.3. Data Preprocessing Before handling the LC-MS datasets in positive and negative mode, each data file was converted to the ABF format. Then, converted data were processed with MS-DIAL v.4.36 [ 54 ] (each mode independently) for peak detection, retention time (RT) shift corrections, peak grouping, and peak alignment across all the samples (parameters are included as Supplementary Tables S4 and S5). The output of data processing is the pairs of MS m/z, LC RT (min), and the peak height of each detected metabolites or metabolite features across all the samples. Missing values were imputed in two rounds using the Random Forest [ 55 ] algorithm included in the “notame” package [56] available in R [57]. 4.6.4. Feature-Clustering In untargeted metabolomics/lipidomics studies, several features can originate from the same metabolite, and thus, they are assumed to be highly correlated. Therefore, we have implemented the feature-clustering algorithm included in the notame [ 56 ] package within the data processing workflow. The algorithm identifies pairs of correlated features within a specified RT window and a correlation threshold (0.1 min and 0.90, respectively). The advantage of this process is that it facilitates the identification of correlated features and generates cleaner datasets, reducing the amount of noise that can disturb the subsequent multivariate analysis. 4.7. Multivariate Statistical Analysis Multivariate exploratory analyses by PCA of the log-transformed and Pareto-scaled values were performed in SIMCA-P (version 15; Umetrics AB, Umeå, Sweden) to discard potential outliers, visualize the total variation of the metabolite profiles, and identify clustering patterns. For models containing one factor, i.e., time or treatment, an Orthogonal Partial Less Squares Discriminant Analysis (OPLS-DA) model was built to identify patterns or metabolic features discriminating between groups. The default seven-round crossvalidation in the SIMCA software package was applied in these discriminant analyses. The cross-validation analysis of variance (CV-ANOVA) was calculated to assess the reliability, and a value ≤ 0.05 was considered significant. Moreover, the R 2 X and Q 2 were evaluated to assess the robustness; in this case, values close to one reflect a reliable model. For comparisons containing two factors, an AMOPLS was performed [ 58 , 59 ]. AMOPLS is a method that integrates the ANOVA-based submatrices associated with each effect and interaction into a single multiblock orthogonal PLS model. In such a way, we can detect any difference on each factor, even if it is subtle. This approach has shown to be very effective in distinguishing between multiple metabolic alterations and facilitating Int. J. Mol. Sci. 2022,23, 3667 17 of 20 interpretation [ 58 ]. As with other standard factorial models, the interpretation can be made using score and loading plots and, in addition, the VIP 2 values, which reflect the contribution of each variable to each main effect or interaction. Moreover, the statistical significance of each effect can be empirically estimated using a permutation test and thus evaluate the validity. 4.8. Univariate Statistical Analysis T-Test For lipidomics feature wise-analysis, a simple Welch’s t-test was performed comparing a pair of classes. Furthermore, the false discovery rate was used to correct for multiple testing and false positives, establishing a cut-off of q < 0.05 and obtaining the t.score for feeding the MS-Peaks to pathways analysis. 4.9. Pathway Analysis with Mummichog and Gene Set Enrichment Analysis (GSEA) Pathway analysis with mummichog v2.06 [ 27 ] and GSEA [ 60 ] was developed using the MS-Peaks-to-Pathway module available on MetaboAnalyst v5.0 [ 28 ]. The input file required m/zvalues, RT values, p-values, t.score, and ionization mode (the mixed option that includes both negative and positive was selected). The molecular weight tolerance was set to 5 ppm and the option to enforce primary ions. Only features with at least one assigned primary ion [M + H] + , [M + Na] + , [M − H 2 O + H] + , [M − H] − , [M − H 2 O − H] − and [M − 2H] 2− were valid. The p-value cut-off was set to α = 0.01 for the mummichog algorithm. We selected the meta-analysis option that combines the results from the GSEA and mummichog. The Lipids Sub-Chemical Class library includes 302 main lipid chemical classes for pathway analysis. Pathways including at least three entries and a combined p-value ≤ 0.05 were considered significantly modulated. It is important noting that in this version of the mummichog algorithm, the RT’s use will help increase the confidence and robustness of the potential compound matches. Retention is used to refine the grouping of signals into empirical compounds. Therefore, there is a reduction in the false-positive annotations and thus an increase in the accuracy of pathway activity prediction. Moreover, the use of the Lipids Sub-Chemical Class library narrows the spectrum of possibilities for annotation and matches with the analytical platform, which focuses on lipids. 5. Conclusions In conclusion, this study confirmed that the FA profile of IVLE, particularly that enriched with ω -3 LC PUFAS, significantly influenced the plasma and erythrocyte functional lipidome. Our findings suggest that changes in specific lipid classes due to high levels of ω -3 LC-PUFA with increases in plasma PC, PE, PI, and PS and paralleled decreases in PEceramides and ceramide-1-phosphate may have an influence on several metabolic pathways and could contribute to improving the inflammatory status of children after HSCT. Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijms23073667/s1. Author Contributions: Conceptualization, A.G. and M.G.-C.; data curation, M.J.d.l.T.-A.; formal analysis, M.J.d.l.T.-A. and M.D.M.; funding acquisition, M.G.-C.; investigation, M.J.d.l.T.-A., K.F.-R., M.A.B.-G. and M.G.-C.; methodology, O.D.R.-H., M.J.d.l.T.-A. and M.D.M.; software, O.D.R.-H.; supervision, M.J.d.l.T.-A. and M.G.-C.; validation, J.L.P.-N., A.G. and M.G.-C.; visualization, M.G.-C.; writing—original draft, O.D.R.-H. and A.G. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by the “Salud Investiga Modalidad Joven 2010” award from the Junta de Andalucía, Spain and CIBERobn. Institutional Review Board Statement: The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Córdoba (Spain) on 18 March 2010. The Clinical Trial Registration Number is NCT02199821. Int. J. Mol. Sci. 2022,23, 3667 18 of 20 Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: Data not presented in the manuscript or supplementary material can be provided if available. Acknowledgments: We thank Rodriguez (Hematology Unit, Hospital Reina Sofía) for her collaboration on the review of the medical records and Maternal-Infant and Developmental Health Network (SAMID), RETICS Carlos III Health Institute (ISCIII), Madrid, Spain (Red SAMID RD12/0022/0003). Conflicts of Interest: The authors declare no conflict of interest. References 1. Chaudhry, M.; Ali, N. Reduced-Intensity Conditioning Hematopoietic Stem Cell Transplantation: Looking Forward to an International Consensus. Blood Res. 2015,50, 69–70. [CrossRef] [PubMed] 2. McGrath, K.H.; Evans, V.; Yap, J. Indications and Patterns of Use for Parenteral Nutrition in Pediatric Oncology. J. Parenter. Enter. Nutr. 2020,44, 632–638. [CrossRef] [PubMed] 3. Evans, J.C.; Hirani, S.P.; Needle, J.J. Nutritional and Post-Transplantation Outcomes of Enteral versus Parenteral Nutrition in Pediatric Hematopoietic Stem Cell Transplantation: A Systematic Review of Randomized and Nonrandomized Studies. Biol. Blood Marrow Transpl. 2019,25, e252–e259. [CrossRef] [PubMed] 4. McMillen, K.K.; Coghlin-Dickson, T.; Adintori, P.A. Optimization of Nutrition Support Practices Early After Hematopoietic Cell Transplantation. Bone Marrow Transpl. 2021,56, 314–326. [CrossRef] 5. Jonas, C.R.; Puckett, A.B.; Jones, D.P.; Griffith, D.P.; Szeszycki, E.E.; Bergman, G.F.; Furr, C.E.; Tyre, C.; Carlson, J.L.; Galloway, J.R.; et al. Plasma Antioxidant Status after High-Dose Chemotherapy: A Randomized Trial of Parenteral Nutrition in Bone Marrow Transplantation Patients. Am. J. Clin. Nutr. 2000,72, 181–189. [CrossRef] 6. Murray, S.M.; Pindoria, S. Nutrition Support for Bone Marrow Transplant Patients. Cochrane Database Syst. Rev. 2017 ,2017, CD002920. [CrossRef] 7. Woods, T.; Tariman, J.; Lee, Y. Enteral and Parenteral Nutrition: An Integrative Literature Review on Nutrition in Pediatric Recipients of Hematopoietic Stem Cell Transplantation. Clin. J. Oncol. Nurs. 2019,23, 351–354. [CrossRef] 8. Nava, T.; Ansari, M.; Dalle, J.H.; de Heredia, C.D.; Güngör, T.; Trigoso, E.; Falkenberg, U.; Bertaina, A.; Gibson, B.; Jarisch, A.; et al. Supportive Care during Pediatric Hematopoietic Stem Cell Transplantation: Beyond Infectious Diseases. A Report from Workshops on Supportive Care of the Pediatric Diseases Working Party (PDWP) of the European Society for Blood and Marrow Transplantation (EBMT). Bone Marrow Transpl. 2020,55, 1126–1136. [CrossRef] 9. Ifversen, M.; Meisel, R.; Sedlacek, P.; Kalwak, K.; Sisinni, L.; Hutt, D.; Lehrnbecher, T.; Balduzzi, A.; Diesch, T.; Jarisch, A.; et al. Supportive Care During Pediatric Hematopoietic Stem Cell Transplantation: Prevention of Infections. A Report from Workshops on Supportive Care of the Paediatric Diseases Working Party (PDWP) of the European Society for Blood and Marrow Transplantation (EBMT). Front. Pediatr. 2021,9, 1126–1136. [CrossRef] 10. Aoyama, T.; Imataki, O.; Mori, K.; Yoshitsugu, K.; Fukaya, M.; Okamura, I.; Enami, T.; Tatara, R.; Ikeda, T.; Okamura-Shiki, I. Nutritional risk in allogeneic stem cell transplantation: Rationale for a tailored nutritional pathway. Ann. Hematol. 2017 ,96, 617–625. [CrossRef] 11. Muscaritoli, M.; Conversano, L.; Torelli, G.F.; Arcese, W.; Capria, S.; Cangiano, C.; Falcone, C.; Fanelli, F.R. Clinical and metabolic effects of different parenteral nutrition regimens in patients undergoing allogeneic bone marrow transplantation. Transplantation 1998,66, 610–616. [CrossRef] [PubMed] 12. Hwang, T.-L.; Chiang, C.-L.; Wang, P.-N. Parenteral nutrition support after bone marrow transplantation: Comparison of total and partial parenteral nutrition during the early posttransplantation period. Nutrition 2001,17, 773–775. [CrossRef] 13. Ren, T.; Cong, L.; Wang, Y.; Tang, Y.; Tian, B.; Lin, X.; Zhang, Y.; Tang, X. Lipid emulsions in parenteral nutrition: Current applications and future developments. Expert Opin. Drug Deliv. 2013,10, 1533–1549. [CrossRef] [PubMed] 14. Manzanares, W.; Langlois, P.L.; Dhaliwal, R.; Lemieux, M.; Heyland, D.K. Intravenous fish oil lipid emulsions in critically ill patients: An updated systematic review and meta-analysis. Crit. Care 2015,19, 167. [CrossRef] 15. Tian, H.; Yao, X.; Zeng, R.; Sun, R.; Tian, H.; Shi, C.; Li, L.; Tian, J.; Yang, K. Safety and efficacy of a new parenteral lipid emulsion (SMOF) for surgical patients: A systematic review and meta-analysis of randomized controlled trials. Nutr. Rev. 2013 ,71, 815–821. [CrossRef] [PubMed] 16. Li, N.-N.; Zhou, Y.; Qin, X.-P.; Chen, Y.; He, D.; Feng, J.-Y.; Wu, X.-T. Does intravenous fish oil benefit patients post-surgery? A meta-analysis of randomised controlled trials. Clin. Nutr. 2014,33, 226–239. [CrossRef] [PubMed] 17. Baena-Gómez, M.; Aguilar, M.D.L.T.; Mesa, M.; Llorente-Cantarero, F.; Navero, J.P.; Gil-Campos, M. Effects of Parenteral Nutrition Formulas on Plasma Lipid Profile in Children with Bone Marrow Transplantation. Ann. Nutr. Metab. 2013 ,63, 103–110. [CrossRef] [PubMed] 18. Baena-Gómez, M.A.; De La Torre Aguilar, M.J.; Mesa, M.D.; Navero, J.L.P.; Gil-Campos, M. Changes in Antioxidant Defense System Using Different Lipid Emulsions in Parenteral Nutrition in Children after Hematopoietic Stem Cell Transplantation. Nutrients 2015,7, 7242–7255. [CrossRef] Int. J. Mol. Sci. 2022,23, 3667 19 of 20 19. Baena-Gómez, M.A.; De La Torre-Aguilar, M.J.; Aguilera-García, C.M.; Olza, J.; Pérez-Navero, J.L.; Gil-Campos, M. Inflammatory Response Using Different Lipid Parenteral Nutrition Formulas in Children After Hematopoietic Stem Cell Transplantation. Nutr. Cancer 2016,68, 804–810. [CrossRef] 20. Fahy, E.; Subramaniam, S.; Brown, H.A.; Glass, C.K.; Merrill, A.H.; Murphy, R.C.; Raetz, C.R.H.; Russell, D.W.; Seyama, Y.; Shaw, W.; et al. A comprehensive classification system for lipids. J. Lipid Res. 2005,46, 839–861. [CrossRef] 21. Han, X. Lipidomics for studying metabolism. Nat. Rev. Endocrinol. 2016,12, 668–679. [CrossRef] [PubMed] 22. Reikvam, H.; Grønningsæter, I.-S.; Mosevoll, K.A.; Lindås, R.; Hatfield, K.; Bruserud, Ø. Patients with Treatment-Requiring Chronic Graft versus Host Disease after Allogeneic Stem Cell Transplantation Have Altered Metabolic Profiles due to the Disease and Immunosuppressive Therapy: Potential Implication for Biomarkers. Front. Immunol. 2018,8, 1979. [CrossRef] [PubMed] 23. Liu, Y.; Huang, A.; Chen, Q.; Chen, X.; Fei, Y.; Zhao, X.; Zhang, W.; Hong, Z.; Zhu, Z.; Yang, J.; et al. A distinct glycerophospholipid metabolism signature of acute graft versus host disease with predictive value. JCI Insight 2019 ,5, e129494. [CrossRef] [PubMed] 24. Michonneau, D.; Latis, E.; Curis, E.; Dubouchet, L.; Ramamoorthy, S.; Ingram, B.; De Latour, R.P.; Robin, M.; De Fontbrune, F.S.; Chevret, S.; et al. Metabolomics analysis of human acute graft-versus-host disease reveals changes in host and microbiota-derived metabolites. Nat. Commun. 2019,10, 5695. [CrossRef] [PubMed] 25. Contaifer, J.D.; Roberts, C.H.; Kumar, N.G.; Natarajan, R.; Fisher, B.J.; Leslie, K.; Reed, J.; Toor, A.A.; Wijesinghe, D.S. A Preliminary Investigation towards the Risk Stratification of Allogeneic Stem Cell Recipients with Respect to the Potential for Development of GVHD via Their Pre-Transplant Plasma Lipid and Metabolic Signature. Cancers 2019,11, 1051. [CrossRef] 26. Kelly, D.L.; Farhadfar, N.; Starkweather, A.; Garrett, T.J.; Yao, Y.; Wingard, J.R.; Mahmud, I.; Menzies, V.; Patel, P.; Alabasi, K.M.; et al. Global Metabolomics in Allogeneic Hematopoietic Cell Transplantation Recipients Discordant for Chronic Graft-versus-Host Disease. Biol. Blood Marrow Transpl. 2020,26, 1803–1810. [CrossRef] 27. Li, S.; Park, Y.; Duraisingham, S.; Strobel, F.H.; Khan, N.; Soltow, Q.A.; Jones, D.P.; Pulendran, B. Predicting Network Activity from High Throughput Metabolomics. PLoS Comput. Biol. 2013,9, e1003123. [CrossRef] 28. Pang, Z.; Chong, J.; Zhou, G.; de Lima Morais, D.A.; Chang, L.; Barrette, M.; Gauthier, C.; Jacques, P.-É.; Li, S.; Xia, J. MetaboAnalyst 5.0: Narrowing the gap between raw spectra and functional insights. Nucleic Acids Res. 2021,49, W388–W396. [CrossRef] 29. Tenenhaus, A.; Philippe, C.; Guillemot, V.; Le Cao, K.-A.; Grill, J.; Frouin, V. Variable selection for generalized canonical correlation analysis. Biostatistics 2014,15, 569–583. [CrossRef] 30. Singh, A.; Shannon, C.P.; Gautier, B.; Rohart, F.; Vacher, M.; Tebbutt, S.J.; LêCao, K.-A. DIABLO: An integrative approach for identifying key molecular drivers from multi-omics assays. Bioinformatics 2019,35, 3055–3062. [CrossRef] 31. Rohart, F.; Gautier, B.; Singh, A.; LêCao, K.-A. mixOmics: An R package for ‘omics feature selection and multiple data integration. PLoS Comput. Biol. 2017,13, e1005752. [CrossRef] [PubMed] 32. González, I.; Cao, K.-A.L.; Davis, M.J.; Déjean, S. Visualising associations between paired ‘omics’ data sets. BioData Min. 2012 , 5, 19. [CrossRef] [PubMed] 33. Kang, J.X.; Wan, J.B.; He, C. Concise Review: Regulation of Stem Cell Proliferation and Differentiation by Essential Fatty Acids and Their Metabolites. Stem Cells 2014,32, 1092–1098. [CrossRef] [PubMed] 34. Bansal, P.; Dahate, P.; Raghuwanshi, S.; Sharma, D.S.; Kovuru, N.; Gutti, U.; Yarla, N.S.; Gutti, R.K.; Narasaiah, K. Current Updates on Role of Lipids in Hematopoiesis. Infect. Disord. Drug Targets 2018,18, 192–198. [CrossRef] [PubMed] 35. Lee, M.K.S.; Al-Sharea, A.; Dragoljevic, D.; Murphy, A.J. Hand of FATe: Lipid Metabolism in Hematopoietic Stem Cells. Curr. Opin. Lipidol. 2018,29, 240–245. [CrossRef] [PubMed] 36. Rodgers, C.; Gregurich, M.A.; Hockenberry, M. Lipid Profiles of Pediatric Hematopoietic Stem Cell Transplant Survivors. J. Pediatr. Oncol. Nurs. 2012,29, 63–69. [CrossRef] [PubMed] 37. Bis, G.; Szlasa, W.; Sondaj, K.; Zendran, I.; Mielcarek-Siedziuk, M.; Barg, E. Lipid Complications after Hematopoietic Stem Cell Transplantation (HSCT) in Pediatric Patients. Nutrients 2020,12, 2500. [CrossRef] 38. Waitzberg, D.L.; Torrinhas, R.S.; Jacintho, T.M. New Parenteral Lipid Emulsions for Clinical Use. J. Parenter. Enter. Nutr. 2006 ,30, 351–367. [CrossRef] 39. Goulet, O.J.; Cai, W.; Seo, J.M. Lipid Emulsion Use in Pediatric Patients Requiring Long-Term Parenteral Nutrition. J. Parenter. Enter. Nutr. 2020,44, S55–S67. [CrossRef] 40. Finn, K.L.; Chung, M.; Rothpletz-Puglia, P.; Byham-Gray, L. Impact of Providing a Combination Lipid Emulsion Compared with a Standard Soybean Oil Lipid Emulsion in Children Receiving Parenteral Nutrition. J. Parenter. Enter. Nutr. 2015 ,39, 656–667. [CrossRef] 41. Kapoor, V.; Malviya, M.N.; Soll, R. Lipid emulsions for parenterally fed term and late preterm infants. Cochrane Database Syst. Rev. 2019,6, CD013171. [CrossRef] [PubMed] 42. Goulet, O.; Lamazière, A.; Nader, E.A.; Talbotec, C.; Wolf, C.; Lambe, C. Erythrocyte fatty acid membrane composition in children on long-term parenteral nutrition enriched with ω-3 fatty acids. Am. J. Clin. Nutr. 2022,115, 422–431. [CrossRef] [PubMed] 43. van Hoogevest, P.; Wendel, A. The Use of Natural and Synthetic Phospholipids as Pharmaceutical Excipients. Eur. J. Lipid Sci. Technol. 2014,116, 1088–1107. [CrossRef] [PubMed] 44. Pironi, L.; Guidetti, M.; Verrastro, O.; Iacona, C.; Agostini, F.; Pazzeschi, C.; Sasdelli, A.S.; Melchiorre, M.; Ferreri, C. Functional lipidomics in patients on home parenteral nutrition: Effect of lipid emulsions. World J. Gastroenterol. 2017 ,23, 4604–4614. [CrossRef] [PubMed] 45. Franco, R.S. Measurement of Red Cell Lifespan and Aging. Transfus. Med. Hemother. 2012,39, 302–307. [CrossRef] [PubMed] Int. J. Mol. Sci. 2022,23, 3667 20 of 20 46. Cabukusta, B.; Nettebrock, N.T.; Kol, M.; Hilderink, A.; Tafesse, F.G.; Holthuis, J.C.M. Ceramide phosphoethanolamine synthase SMSr is a target of caspase-6 during apoptotic cell death. Biosci. Rep. 2017,37, BSR20170867. [CrossRef] 47. Schütze, S.; Potthoff, K.; Machleidt, T.; Berkovic, D.; Wiegmann, K.; Krönke, M. TNF activates NFκ B by phosphatidylcholinespecific phospholipase C-induced “Acidic” sphingomyelin breakdown. Cell 1992,71, 765–776. [CrossRef] 48. Aureli, M.; Murdica, V.; Loberto, N.; Samarani, M.; Prinetti, A.; Bassi, R.; Sonnino, S. Exploring the link between ceramide and ionizing radiation. Glycoconj. J. 2014,31, 449–459. [CrossRef] 49. Kallunki, T.; Olsen, O.D.; Jäättelä, M. Cancer-associated lysosomal changes: Friends or foes? Oncogene 2013 ,32, 1995–2004. [CrossRef] 50. Skoczen, S.; Tomasik, P.J.; Fijorek, K.; Strojny, W.; Wieczorek, A.; Balwierz, W.; Sztefko, K.; Siedlar, M. Concentrations of adipokines in children before and after hematopoietic stem cell transplantation. Pediatr. Hematol. Oncol. 2016,33, 21–38. [CrossRef] 51. Yuan, Y.; Zhao, J.; Li, T.; Ji, Z.; Xin, Y.; Zhang, S.; Qin, F.; Zhao, L. Integrative metabolic profile of myelodysplastic syndrome based on UHPLC–MS. Biomed. Chromatogr. 2021,35, e5136. [CrossRef] [PubMed] 52. Lepage, G.; Roy, C.C. Direct transesterification of all classes of lipids in a one-step reaction. J. Lipid Res. 1986 ,27, 114–120. [CrossRef] 53. Gil-Campos, M.; Larqué, E.; Ramírez-Tortosa, M.C.; Linde, J.; Villada, I.; Cañete, R.; Gil, A. Changes in plasma fatty acid composition after intake of a standardised breakfast in prepubertal obese children. Br. J. Nutr. 2007 ,99, 909–917. [CrossRef] [PubMed] 54. Tsugawa, H.; Cajka, T.; Kind, T.; Ma, Y.; Higgins, B.; Ikeda, K.; Kanazawa, M.; VanderGheynst, J.; Fiehn, O.; Arita, M. MS-DIAL: Data-independent MS/MS deconvolution for comprehensive metabolome analysis. Nat. Methods 2015,12, 523–526. [CrossRef] 55. Gromski, P.S.; Xu, Y.; Kotze, H.L.; Correa, E.; Ellis, D.I.; Armitage, E.G.; Turner, M.L.; Goodacre, R. Influence of Missing Values Substitutes on Multivariate Analysis of Metabolomics Data. Metabolites 2014,4, 433–452. [CrossRef] [PubMed] 56. Klåvus, A.; Kokla, M.; Noerman, S.; Koistinen, V.M.; Tuomainen, M.; Zarei, I.; Meuronen, T.; Häkkinen, M.R.; Rummukainen, S.; Babu, A.F.; et al. “Notame”: Workflow for Non-Targeted LC–MS Metabolic Profiling. Metabolites 2020,10, 135. [CrossRef] 57. R Core Development Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing: Vienna, Austria, 2012. 58. González-Ruiz, V.; Schvartz, D.; Sandström, J.; Pezzatti, J.; Jeanneret, F.; Tonoli, D.; Boccard, J.; Monnet-Tschudi, F.; Sanchez, J.-C.; Rudaz, S. An Integrative Multi-Omics Workflow to Address Multifactorial Toxicology Experiments. Metabolites 2019 ,9, 79. [CrossRef] [PubMed] 59. Boccard, J.; Rudaz, S. Exploring Omics data from designed experiments using analysis of variance multiblock Orthogonal Partial Least Squares. Anal. Chim. Acta 2016,920, 18–28. [CrossRef] 60. Subramanian, A.; Tamayo, P.; Mootha, V.K.; Mukherjee, S.; Ebert, B.L.; Gillette, M.A.; Paulovich, A.; Pomeroy, S.L.; Golub, T.R.; Lander, E.S.; et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 2005,102, 15545–15550. [CrossRef]