Parenteral lipid emulsions induce unique ileal fatty acid and metabolomic profiles but do not increase the risk of necrotizing enterocolitis in preterm pigs

Abstract

Necrotizing enterocolitis (NEC) is a manifestation of maladaptive intestinal responses in preterm infants centrally medicated by unattenuated inflammation. Early in the postnatal period, preterm infants develop a deficit in arachidonic and docosahexaenoic acid, both potent regulators of inflammation. We hypothesized that the fatty acid composition of parenteral lipid emulsions uniquely induces blood and intestinal fatty acid profiles which, in turn, modifies the risk of NEC development. Forty-two preterm pigs were randomized to receive one of three lipid emulsions containing 100% soybean oil (SO), 15% fish oil (MO15), or 100% fish oil (FO100) with enteral feedings over an 8-day protocol. Blood and distal ileum tissue were collected for fatty acid analysis. The distal ileum underwent histologic, proteomic, and metabolomic analyses. Eight pigs [3/14 SO (21%), 3/14 MO15 (21%), and 2/14 FO100 (14%)] developed NEC. No differences in NEC risk were evident between groups despite differences in induced fatty acid profiles in blood and ileal tissue. Metabolomic analysis of NEC versus no NEC tissue revealed differences in tryptophan metabolism and arachidonic acid-containing glycerophospholipids. Proteomic analysis demonstrated no differences by lipid group; however, 15 proteins differentiated NEC versus no NEC in the domains of tissue injury, glucose uptake, and chemokine signaling. Exposure to parenteral lipid emulsions induces unique intestinal fatty acid and metabolomic profiles; however, these profiles are not linked to a difference in NEC development. Metabolomic and proteomic analyses of NEC versus no NEC intestinal tissue provide mechanistic insights into the pathogenesis of NEC in preterm infants.

NEW & NOTEWORTHY Exposure to parenteral lipid emulsions induces unique intestinal fatty acid and metabolomic profiles; however, these profiles are not linked to a difference in NEC risk in preterm pigs. Metabolomic and proteomic analyses provide mechanistic insights into NEC pathogenesis. Compared with healthy ileal tissue, metabolites in tryptophan metabolism and arachidonic acid-containing glycerophospholipids are increased in NEC tissue. Proteomic analysis differentiates NEC versus no NEC in the domains of tissue injury, glucose uptake, and chemokine signaling.

INTRODUCTION

Necrotizing enterocolitis (NEC) is the extreme manifestation of a maladaptive intestinal response in preterm infants often after the introduction of enteral feedings (1). Although the specific causes for the pathogenesis of NEC remain unclear, several risk factors have been shown to be associated with the development of NEC, including prematurity, low birth weight, enteral feeding, bacterial colonization, and ischemia/reperfusion (2). Although infants may uniquely possess a combination of risk factors that lead to NEC, expression of disease is mediated through a final common pathway of dysregulated inflammation (3, 4). Many regulators and switch points exist in the regulation of inflammation and one pathway of interest in the preterm infant are those that are mediated by long-chain polyunsaturated fatty acids (LCPUFAs).

Accretion of LCPUFAs in utero from the maternal to fetal circulation and in tissues occurs largely during the last trimester. Preterm infants miss much of the last trimester in utero and have insufficient time to accrue adequate adipose tissue reservoirs to maintain the high birth levels of LCPUFAs such as docosahexaenoic acid (DHA) and arachidonic acid (AA) necessary for optimal fetal development. As a result of the lack of adipose tissue reservoirs coupled with inadequate postnatal nutritional delivery, the preterm infant rapidly develops systemic deficits in DHA and AA. It is known that DHA and AA play critical roles in the development of neural tissues such as the brain and eye, whereas accumulating evidence demonstrates its organogenetic effects to other organs such as the lung and gut (5, 6). Postnatal fatty acid alterations contribute to the pathophysiology of major morbidities in the preterm infant (7). These clinical associations may be due to not only altered organogenesis but also dysregulated inflammation (4).

Currently available parenteral lipid emulsions have varying compositions from emulsions that are 100% soybean oil (SO)-based to lipid emulsions containing other lipid sources such fish oil, either as a part of a mixed composition or as a pure (100%) fish oil lipid emulsion. The high omega-6 (n-6) fatty acid and phytosterol content of 100% soybean-oil-based lipid emulsions have been associated with increased oxidative stress, inflammation, and cholestasis in preterm infants (8). A 100% fish oil (FO100)-based lipid emulsion has been shown to reverse parenteral nutrition (PN)-associated cholestasis (PNAC) (8–10). In addition, enteral DHA supplementation in animal models and small clinical trials in preterm infants have shown to reduce diseases of prematurity such as chronic lung injury, retinopathy of prematurity, and NEC (10–13). Although the role of LCPUFAs in modulating inflammation has been well characterized, the biological priming effect of early PN in altering systemic and tissue LCPUFA levels, -omic profiles, and subsequent disease risk is largely understudied.

In this study, we hypothesized that induced changes in systemic and intestinal fatty acids by parenteral lipid emulsions will be linked to priming of altered downstream expression of proteins and metabolites that influence the risk of developing NEC. We utilized both proteomic and metabolomic approaches to characterize biomarker changes associated with the development of NEC using a preterm piglet model. To evaluate our overall hypothesis, our study objectives were to 1) characterize the changes in circulating fatty acid levels; intestinal morphology and fatty acid levels; and, intestinal tissue proteomic and metabolomic profiles induced by administering specific lipid emulsions with varying LCPUFA concentrations, and 2) to conduct bioinformatic analyses of multi-omic data to determine the risk of NEC secondary to specific proteomic and metabolomic signatures induced by parenteral lipid emulsions.

MATERIALS AND METHODS

Animal Care

All studies were approved by the Animal Care and Use Committee at Baylor College of Medicine (Houston, TX). Five litters from five sows were used in this study. Timed-pregnant sows were obtained from a commercial swine farm, housed at the animal facility of the Children’s Nutrition Research Center with food and water provided ad libitum. Preterm pigs were delivered by cesarean section at 103 days of gestation (114 days’ term) and placed in acrylic incubators housed at 32°C as described previously (14, 15). After delivery, pigs underwent surgery for placement of jugular catheters for PN administration and orogastric (OG) feeding tubes and PN was initiated following line placement. During the first 24 h, maternal plasma (16 mL/kg in three doses) was administered intravenously for passive immunological protection. Postoperatively, pigs were monitored continuously over the 8-day study period, including rectal temperatures and vital signs twice daily. Nutrition rates were adjusted according to body weights taken every other day.

Nutrition Protocol, Sample Collection, and Growth Monitoring

Following surgery on day 0, pigs received PN at 5.5 mL/kg/h (132 mL/kg/day or 50% of total daily requirement). At this rate, PN provided (per kg/day) was as follows: 410 kJ energy, 12.5 g dextrose, 6.5 g amino acids, and 2.5 g fat. At this time, pigs were randomized to receive one of three lipid emulsions (all from Fresenius Kabi, Bad Homburg, Germany): 1) SO, soybean oil (intralipid 20%, n = 14); 2) MO15, mixed oil emulsion with 15% fish oil (SMOFlipid 20%, n = 14); 3) FO100, 100% fish oil (Omegaven 10%, n = 14). The fatty acid, vitamin, and phytosterol content of these lipid emulsions have been previously summarized by Raman et al. (16). These lipid emulsions were not added to the PN-bags but infused separately via syringe pump before joining the jugular catheter using a Y-connector. Because Omegaven was provided as a 10% lipid emulsion, both, Intralipid and SMOFlipid (20% lipid emulsions) were diluted 1:1 with sterile water to administer equal volumes of PN to all pigs. Because of this dilution, the PN infusion rate at 50% of full intake was increased by 0.5 mL/kg/h. On days 1 and 2, PN was successively increased to 70% and 90% of full intake. Enteral nutrition (EN), administered at ambient temperature (30–32°C) via orogastric tube in 3-h intervals, was initiated on day 3 at 10% of full intake and successively increased to 50% until day 8, whereas PN was decreased such that pigs received their full nutritional requirement by a combination of enteral nutrition and PN on days 3–8. Enteral nutrition consisted of infant formula (Pepdite Junior, Nutricia North America, Gaithersburg, MD) with added whey protein isolate (NOW Foods, Bloomingdale, IL) and medium-chain triacylglyceride (MCT) oil (NOW Foods, Bloomingdale, IL) and provided energy and macronutrient levels comparable to sow’s milk (17). At 50% intake (day 7 and day 8), enteral formula provided (kg/day) 487 kJ energy, 7.6 g protein (mainly whey protein isolate), 5.9 g carbohydrate (corn syrup solids), 7.3 g fat (mainly MCT oil), and minerals and vitamins. Rectal temperatures, oxygen saturation, pulse, as well as signs of distress (i.e., emesis, blood oxygen saturation, bloody diarrhea, and distension), were monitored closely throughout the study. Pigs were euthanized at the end of the 8-day protocol or when there was strong clinical suspicion that a pig had developed NEC (e.g., abdominal distension, bloody diarrhea, fever, emesis, low oxygen saturation, etc.) (14, 18). Blood samples (0.75 mL each) collected via jugular catheter into EDTA tubes on days 0, 3, 6, and 8 were processed to separate plasma and red blood cells. Both fractions were frozen in separate tubes and stored at −80°C. Samples of distal ileum were collected and either frozen and stored at −80°C or fixed in 10% formalin. Pigs were weighed daily, and their intake adjusted accordingly. Growth velocity (g/kg/day) was calculated by [(final weight, g − birth weight, g)/(birth weight, kg)/8].

Red Blood Cell Membrane and Tissue Fatty Acid Levels

Fatty acids were isolated and methylated using a modified Folch method as previously described (7, 19). Red blood cell and distal ileum samples were quantified by gas chromatography-mass spectroscopy (GC-MS) with a Supelcowax-10 column (Sigma-Aldrich). Peak identification was based upon comparison of both retention time and mass spectra of the unknown peak to that of known standards. Fatty acid methylated ester (FAME) mass was determined by comparing areas of unknown FAMEs to that of a fixed concentration of 17:0 internal standard. Response factors were determined for each individual FAME to correct for GC-MS total ion chromatogram discrepancies in quantification. These factors were determined using a GLC reference standard which contained known masses of FAMEs ranging from 14 to 24°C. The response ratio of each FAME is corrected to a fixed amount ratio for each FAME relative to 17:0. Individual FA is expressed as a percent of the total FA mass (mol%).

NEC Scoring and Ileum Morphometry

The presence of NEC and graded severity was assessed by histological NEC scoring performed on hematoxylin-eosin-stained sections of proximal jejunum, distal ileum, and colon. The severity of intestinal injury was graded by two independent observers using an intestinal injury score (0–4). Histological changes were graded as follows; score 0, intact villi with no damage; score 1, epithelial cell lifting or separation of lamina propria; score 2, partial loss of villi; score 3, complete loss of villi; score 4, transmural necrosis (20). A histological NEC severity score of ≥2 was considered positive for intestinal injury/NEC. A histological score was assigned to each specimen based on the area with the worst injury. Height of intact villi and depth of crypts were measured using Aperio ImageScope software (Leica Biosystems, Buffalo Grove, IL). For every section, at least five randomly selected areas were counted. Villus height was measured in continuously visible villi that ended with a rounded tip using a ×20 objective from the tip of villous to the entrance of the crypt opening. Crypt depth was measured from crypt base to villus-crypt junction.

Ileum Proteomic and Metabolomic Measurements

Proteomic analysis on distal ileum tissue was performed at BIDMC Genomics, Proteomics, Bioinformatics and Systems Biology Center as previously described (21, 22) using the SOMAscan Assay Kit 1.3k, Cells and Tissue (Cat. No. 900-00009), following the manufacturer’s recommended protocol. This assay measures 1,305 human proteins and has good cross reactivity with other mammalian species, though cross reactivity with porcine proteins has not been specifically measured. Distal ileum tissue was homogenized in T-PER (Thermo Scientific, Cat. No. 78510) containing 1× Halt protease inhibitor (Thermo Scientific, Cat. No. 78430) to extract the proteins. In total, 2.4 µg of tissue lysate protein per piglet was used to measure all 1,305 proteins in a multiplex manner. Three controls provided by the kit and one nonprotein buffer control were run in parallel with samples per plate. Median normalization and calibration were performed according to standard quality control (QC) protocols at SomaLogic.

Metabolomic analyses of distal ileum tissue were analyzed by Metabolon (Durham, NC). Each sample was accessioned into the Metabolon laboratory information management system (LIMS) and assigned a unique identifier. All samples were prepared using the automated MicroLab STAR system from Hamilton Company. The liquid chromatography-mass spectrometry (LC-MS) portion of the platform was based on a Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer operated at 35,000 mass resolutions. The informatics system consisted of four major components: LIMS, data extraction and peak identification software, data processing tools for QC, and data interpretation. Raw data were extracted, peak identified, and QC processed using Metabolon’s hardware and software. Compounds were identified by comparison to library entries of purified standard or recurrent unknown entities. Peaks were quantified using area-under-the-curve.

Statistical Analyses

All reported measures were evaluated for normality. Normally distributed measures were reported as means ± SD, whereas non-normal measures as median ± interquartile range (Q1–Q3). Mann–Whitney U test was used to compare morphometric differences between NEC and no NEC groups. Kruskal–Wallis test was used to compare differences in vital signs, morphometrics, and red blood cell fatty acid levels by lipid groups. When significant, a two-way ANOVA using generalized linear modeling following rank-normal transformation of data was used to compare unbalanced, repeated measures over the duration of the protocol. Differences in proteomic and metabolomic profiles between NEC and no NEC groups were determined using a median fold change. To account for lipid differences in NEC phenotype, a Kruskal–Wallis test was used to evaluate differences in proteins and metabolites by lipid groups. If significant, as in the case of metabolites, a two-way ANOVA was performed using rank-normalized metabolomic data with lipid groups (SO, MO15, FO100) and NEC incidence (no NEC vs. NEC) as main effects. All P values generated from multiple comparisons were adjusted with Benjamini–Hochberg false discovery rate (FDR). A median fold change cutoff of ≥|1.5| was considered significant change for SOMAscan proteins and has been validated by ELISA as previously described (23, 24). A median fold change of ≥|2.0| was used for metabolomic analyses. All P values were adjusted by FDR and q < 0.05 was considered statistically significant.

Data analyses were performed using R software (version 3.5.2, R Core Team 2018a) within RStudio (Version 1.1.453, RStudio) using the tidyverse (25), ggplot2 (25), phyloseq (26), and multcomp (27) packages. Heatmaps were generated using pheatmap (28) package. For all other analyses, statistical significance was considered at P < 0.05.

RESULTS

Piglet Cohort Characteristics

The total number of live animals delivered was 49. One animal died before surgery, five animals died early due to immaturity/respiratory insufficiency, and one animal accidentally died due to lipid syringe malfunction. Thus, a total of 42 pigs from five litters were included in this study (Table 1). Eight pigs (3/14 in SO, 3/14 in MO15, and 2/14 in FO100) developed NEC over the 8-day protocol. Overall, no significant difference was observed in birth weight and postnatal growth velocity between lipid groups. Comparisons of vital signs (temperature, oxygen saturation, heart rate) in healthy (no NEC) lipid groups and between NEC and no NEC groups are shown in Supplemental Fig. S1 (all Supplemental material is available at https://doi.org/10.6084/m9.figshare.13173191). No significant differences in vital signs were observed in healthy pigs within lipid groups (Supplemental Fig. S1A). However, compared with healthy pigs with no NEC, pigs with NEC showed significantly lower median oxygen saturation (%) at day 4 (98.0% vs. 93.5%, P = 0.01) and day 5 (98.5% vs. 88.0%, P = 0.002) (Supplemental Fig. S1B).

Table 1.

Piglet cohort characteristics by lipid groups, median (IQR)

Lipid	Litter	n	% Male	Birth Weight Grams (IQR)	Growth Velocity, g/kg/day (IQR)	Mortality Due to NEC (n, %)	NEC Mortality Day (range)
Soybean oil (SO)	Total	14	50.0	901 (277)	60.4 (12.1)	3 (21.4)	5 (3–7)
	Litter 1	4		1125 (228)	53.7 (4.46)	2
	Litter 2	2		873 (67)	65.8 (5.47)	1
	Litter 3	0		-	-	0
	Litter 4	4		941 (240)	66.6 (4.10)	0
	Litter 5	4		778 (151)	59.4 (8.50)	0
15% Fish oil (FO15)	Total	14	57.1	911 (274)	61.2 (20.8)	3 (21.4)	4 (1–5)
	Litter 1	3		1104 (145)	59.0 (18.9)	0
	Litter 2	3		870 (178)	37.6 (14.7)	2
	Litter 3	1		1004 (0.0)	69.6 (0.0)	0
	Litter 4	4		1054 (193)	62.4 (7.04)	1
	Litter 5	3		783 (94)	71.8 (9.05)	0
100% Fish oil (FO100)	Total	14	50.0	899 (342)	68.0 (14.3)	2 (14.3)	4.5 (3–6)
	Litter 1	5		1157 (374)	57.1 (13.2)	0
	Litter 2	2		842 (113)	81.7 (9.18)	1
	Litter 3	1		699 (0.0)	43.1 (0.0)	0
	Litter 4	3		1069 (163)	70.4 (5.49)	0
	Litter 5	3		783 (59)	62.4 (5.44)	1

Medians within each group without a common letter are statistically different, P < 0.05.

IQR, interquartile range; NEC, necrotizing enterocolitis.

Intestinal Morphometry

Villus height, crypt depth, and ratio of villus height to crypt depth were quantified using H&E stained distal ileal sections (Fig. 1). Pigs from the healthy no NEC group displayed normal intestinal morphology, intact architecture of intestinal epithelium, intact long villi, and well-organized crypts at the base of villus with no differences between the lipid groups (Fig. 1, B–D). In contrast, pigs in the NEC group had extensive evidence of villus disruption as shown by partial or complete loss of villi and separation of lamina propria (Fig. 2). Median villus height was significantly reduced in NEC group compared with healthy no NEC (371.0 ± 96.9 vs. 546.7 ± 185.4 µm, P = 0.003; Fig. 3B) with no difference in crypt depth (Fig. 2C). Villus height to crypt depth ratio was also significantly reduced in the NEC group compared with no NEC (3.7 ± 0.8 vs. 5.5 ± 1.3, P = 0.002; Fig. 2D).

Figure 1.

Effect of various lipids on intestinal morphology in pigs. Representative images (at ×20) of hematoxylin-eosin (H&E)-stained sections of distal ileal segments of the small intestine are shown for healthy pigs from various lipid groups, soybean oil (SO, n = 10), mixed oil with 15% fish oil (MO15, n = 11), 100 percent fish oil (FO100, n = 11). Four necrotizing enterocolitis (NEC) cases showed complete villus loss and are not included in (B) and (D). A: results are shown for villus height (B), crypt depth (C), and ratio of villus height to crypt depth (D) as box plots (median ± IQR). n represents the number of pigs in each group. Scale bar = 100 μm.

Figure 2.

Small intestinal morphology in distal ileum of piglet from healthy no necrotizing enterocolitis (NEC) (n = 28) and NEC (n = 8) groups. Four NEC cases showed complete villus loss and are not included in B and D. Representative images (at ×20) of hematoxylin-eosin (H&E)-stained sections of distal ileal segments of the small intestine are shown for pigs from healthy no NEC and NEC group (A). Results are shown for villus height (B), crypt depth (C), and ratio of villus height to crypt depth (D) as box plots (median ± IQR) and labeled points without a common letter are significantly different, P < 0.05. Scale bar = 100 μm.

Figure 3.

Specific n-3 fatty acid profiles in the red blood cell (RBC) membranes (mol%) across postnatal age as a function of lipid groups. Concentration of fatty acids was determined by gas chromatography-mass spectroscopy (GC-MS) and are presented as mol% in the form of boxplots. Labeled points without a common letter represent a statistically significant difference of P < 0.05 for each postnatal age. The number of pigs in each group is as follows: day 0 (soybean oil, SO = 14; mixed oil with15% fish oil, MO15 = 14; 100 percent fish oil, FO100 = 14); day 3 (SO = 14, MO15 = 13, FO100 = 14); day 6 (SO = 12, MO15 = 11, FO100 = 12); day 8 (SO = 11, MO15 = 11, FO100 = 12). ALA, α-linolenic acid; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid.

Red Blood Cell and Distal Ileum Fatty Acids

Specific n-3 [α-linolenic acid (ALA), eicosapentaenoic acid (EPA), DHA] and n-6 [linoleic acid (LA), dihomo-γ linolenic acid (DGLA), AA] fatty acid levels in red blood cells are shown in Figs. 3 and 4, respectively. Although birth levels of n-3 fatty acids did not differ between lipid groups, levels of EPA and DHA at days 3, 6, and 8 were significantly higher in FO100 group compared with MO15 and SO lipid groups (Fig. 3 and Supplemental Table S1). Compared with birth levels, the highest change in n-3 fatty acids occurred in FO100 group at day 3, with significant increase in both EPA (0.04 ± 0.03 to 2.8 ± 0.4 mol%, P < 0.01) and DHA (1.9 ± 0.5 to 3.3 ± 0.4 mol%, P < 0.01). No significant change was observed in MO15 and SO lipid groups. RBC membrane n-6 FA LA, a precursor of n-6 LCPUFAs, was significantly highest in the SO group and lowest in the FO100 group at days 3, 6, and 8 (Fig. 4 and Supplemental Table S1). Levels of the n-6 LCPUFA AA were maintained in both the SO and MO15 groups from birth to day 6, but significantly decreased from birth to day 6 in FO100 group.

Figure 4.

Specific n-6 fatty acid profiles in the red blood cell (RBC) membranes (mol%) across postnatal age as a function of lipid groups. Concentration of fatty acids was determined by gas chromatography-mass spectroscopy (GS-MS) and are presented as mol% in the form of boxplots. Labeled points without a common letter represent a statistically significant difference of P < 0.05 for each postnatal age. The number of pigs in each group is as follows: day 0 (soybean oil, SO = 14; mixed oil with 15% fish oil; MO15 = 14; 100 percent fish oil, FO100 = 14); day 3 (SO = 14, MO15 = 13, FO100 = 14); day 6 (SO = 12, MO15 = 11, FO100 = 12); day 8 (SO = 11, MO15 = 11, FO100 = 12). AA, arachidonic acid; DGLA, dihomo-γ linolenic acid; LA, linoleic acid.

Changes in ileal tissue levels of AA, EPA, and DHA at postnatal day 8 are shown in Fig. 5 and Supplemental Table S2. Similar to the RBC membrane data, levels of AA in the ileum were significantly lower in the FO100 group (6.0 ± 0.5 mol%, P < 0.001) compared with MO15 (9.3 ± 1.3 mol%) and SO (10.5 ± 2.0 mol%) groups, but no difference was seen between the SO and MO15 groups (P = 0.3). In contrast, EPA levels were significantly highest in the FO100 group (6.6 ± 1.0 mol%), followed by MO15 (1.5 ± 0.2 mol%) and SO (0.2 ± 0.1 mol%) groups. DHA levels were significantly highest in the FO100 (5.9 ± 0.9 mol%) group, followed by MO15 (3.8 ± 0.8 mol%) and SO (1.9 ± 1.0 mol%) groups.

Figure 5.

Fatty acid profiles in the distal ileum tissue (mol%) as a function of the lipid group. The number of pigs in each group is as follows: soybean oil (SO, n = 11), mixed oil with 15% fish oil (MO15, n = 11), 100 percent fish oil (FO100, n = 12). Concentration of critical n-3 and n-6 fatty acids was determined by gas chromatography-mass spectroscopy (GS-MS) and are presented as mol% in the form of boxplots. Labeled points without a common letter represent a statistically significant difference of P < 0.05 for each postnatal age. AA, arachidonic acid; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid.

Distal Ileum Proteomics and Metabolomics

A total of 1,305 proteins in distal ileum tissue were measured by the SOMAscan Assay. No significant differences in protein expression were observed across lipid groups. However, a total of 15 proteins that were significantly different (P < 0.05, ≥|1.5| fold change) between NEC and no NEC groups and their relative expressions are shown as a heatmap in Fig. 6. Notably, proteins implicated in angiogenesis and tissue injury (e.g., thrombospondin-1, stanniocalcin-1) were increased in ileal tissue of NEC pigs. In contrast, proteins implicated in glucose metabolism (e.g., glucokinase regulatory protein, GKRP; 5′-adenosine monophosphate-activated protein kinase, AMPK) and chemokine signaling (e.g., CAMK2A, 2B, 2D; MAPK8, 14) were significantly decreased in ileal tissue of NEC compared with no NEC pigs.

Figure 6.

Analysis of protein expression profiles in distal ileum tissue clustered by relative expression in pigs. The heatmap shows the relative expression of proteins that are significant different (false discovery rate, FDR-adjusted q < 0.05, fold change ≥ |1.5|) between no necrotizing enterocolitis (NEC) (n = 21) and NEC (n = 7) pigs.

After excluding unknown and xenobiotic compounds, and metabolites with ≥15% missing values, a total of 644 unique metabolites were analyzed. Among healthy pigs without NEC, metabolomic profile significantly varied by lipid groups. Relative to SO, metabolite levels in FO100 showed increased bile acid metabolites (e.g., 6-oxolithocholate, 7-ketolithocholate), increased glutathione metabolites (e.g., glutathione, reduced), and decreased AA-containing glycerophospholipid metabolites, whereas no significant difference in metabolomic levels was found between SO and MO15 (Fig. 7). Comparing NEC with the no NEC groups, 174 metabolites were significantly different by two-way ANOVA adjusting for lipid groups (FDR-adjusted q < 0.05, ≥|2.0| fold change, Fig. 8). Of the significant metabolites, 77 (44%) were lipids and 45 (26%) were amino acids. Overall, four main metabolic pathways were identified to be implicated in NEC compared with healthy no NEC groups: amino acids, carbohydrate, cell-to-cell signaling, and lipid metabolism (Fig. 9). Notably, the amino acids serotonin, glutamate, and kynurenine were significantly increased in NEC. Similarly, n-6 LCPUFA metabolites such as DGLA (20:3n6) and adrenate (22:4:n6), AA-containing glycerophospholipids (e.g., 1-arachidonoyl-GPE), and prostaglandins (e.g., prostaglandin A2) were also significantly increased in NEC compared with healthy no NEC groups.

Figure 7.

Heatmap of metabolomic profiles of healthy [no necrotizing enterocolitis (NEC)] pigs clustered by lipid emulsion groups, soybean oil (SO, n = 11), mixed oil with 15% fish oil (MO15, n = 11), and 100 percent fish oil (FO100, n = 12). The heatmap shows the relative expression of the top 50 significantly different metabolites between lipid groups by one-way ANOVA.

Figure 8.

Heatmap of metabolomic profiles clustered by the relative expression in pigs. The heatmap shows the relative expression of top 50 metabolites that are significant different (false discovery Rate, FDR-adjusted q < 0.05, fold change ≥ |2.0|) between no necrotizing enterocolitis (NEC) (n = 34) and NEC (n = 8) pigs by two-way ANOVA (lipid group × NEC incidence).

Figure 9.

Summary of significant proteins and metabolites between necrotizing enterocolitis (NEC) and no NEC in the distal ileum of preterm pigs. Subpathways associated with proteins and metabolites were identified using Ingenuity, DAVID, and Panther pathway analysis tools. Statistical significance was determined by a |1.5| median fold change for proteins and |2.0| median fold change for metabolites with FDR-adjusted P < 0.05.

DISCUSSION

The increasing availability and use of parenteral lipid emulsions with varying amounts of other oil sources including fish oil in the preterm infant highlight the importance of interrogating induced tissue responses as a function of parenteral lipids. We utilized a large animal model to study these interactions as murine models do not allow for continuous parenteral infusions in the early postnatal period. Although NEC risk was not influenced by lipid emulsion type, we were able to demonstrate that lipid type does influence intestinal fatty acid levels and metabolomics. In addition, NEC, regardless of lipid type, is associated with specific proteomic and metabolomic profiles within ileal tissue that are implicated in amino acid, carbohydrate, lipid, and cell-to-cell signaling pathways. We also confirmed similar findings that have been associated with NEC pathogenesis in human infants, such as decreased oxygen saturation and extensive villus disruption in the ileum (29, 30), thus providing further insight into other biomarker changes associated with NEC.

Fatty acid levels in red blood cells and intestinal tissue reflected the relative amounts of fatty acids present in lipid emulsions administered during the 8-day protocol. The biogenesis of n-3 and n-6 LCPUFAs relies on downstream conversion of the essential fatty acids linoleic acid (LA) and alpha-linolenic acid (ALA) by desaturase and elongase enzymes. Desaturase activity is reported to be limited in infants (31, 32), and this reduces the formation of downstream LCPUFAs. As a result, DHA and AA are considered conditionally essential and require supplementation to meet recommended daily intakes (33–35). Our results showed that high LA in the SO group, as well as high EPA and DHA supplementation in the fish oil groups rapidly increased their respective fatty acid levels in RBC membranes and in the distal ileum. It was also noted that RBC membrane levels of AA were maintained in the SO lipid group from birth to day 8. In contrast, in the FO100 group, RBC membrane AA levels significantly decreased over time and were significantly lower versus the SO group at day 6 and versus both the SO and the MO15 group at day 8 despite this lipid emulsion containing higher AA levels compared with the other lipid emulsions evaluated in this study. Similarly, in the MO15 group, RBC membrane AA levels significantly decreased over time and were significantly lower versus the SO group at day 8. This decrease occurred despite the higher in AA content of the FO100 compared with the other emulsions evaluated in this study. This can be explained by the lower content of the precursor fatty acid, LA, in the FO 100 and MO15 emulsions or due to a reduction of AA synthesis in the presence of high EPA levels.

Proteomic changes observed in the NEC pigs reflected markers of tissue injury and decreased glucose sensing and absorption in the intestinal tissue. Increased expression of stanniocalcin-1, which has been shown to be involved in re-epithelialization of tissue during wound healing (36), and thrombospondin-1, which plays a role in gastrointestinal inflammation through neutrophil-mediated phagocytosis of damaged cells during injury (37, 38), reflects compensatory mechanisms in response to disrupted intestinal morphometry in NEC pigs. Decreased expression of the energy sensors, such as GKRP and AMPK, as well as chemokine signaling kinases (CAMKs and MAPKs) contribute to decreased glucose uptake in the NEC ileum. AMPK facilitates glucose absorption by increasing the expression of glucose transporters in the apical membrane (39, 40). In animal models, AMPK has been shown to modulate intestinal inflammation by reducing proinflammatory cytokine production in intestinal macrophages (41–43). Inhibition of AMPK has also been shown to elevate the proinflammatory LPS-induced cytokines TNF-α, IL-6, and COX-2 in macrophages (44). In addition, preterm piglet studies have also shown that poor glucose uptake increases the severity of NEC (45, 46). Therefore, decreased expression of the AMPK signaling pathway in intestinal tissue of NEC pigs reflects a limitation in glucose absorption by decreasing glucose transport, and potentially increases proinflammatory mediators that increase inflammation observed in pigs with NEC.

Changes in amino acid metabolism, particularly tryptophan metabolism, may also contribute to intestinal inflammation in NEC. The essential amino acid tryptophan is metabolized in the gut mainly by the kynurenine and serotonin (5-HT) pathways (47). Our results showed increased tryptophan catabolism evidenced by increased serotonin (17.2-fold) and kynurenine (2.3-fold) metabolites in the intestinal tissue. Increasing evidence elucidates an association between serotonin and gastrointestinal inflammation. In animal models, prolonging the effects of serotonin either through genetic deletion of serotonin transporter (SERT) in a NEC model (48), or blocking serotonin reuptake in a colitis model (49) increased intestinal inflammation. Although the exact mechanism underlining serotonin activation of proinflammatory cytokines is still being explored, a few studies have shown that serotonin binds to serotonergic receptors on immune cells through activation of the NF-κB signaling pathway, stimulating the production of proinflammatory cytokines in the gut (50, 51).

A significant increase in the levels of n-6 LCPUFAs and proinflammatory lipid metabolites, AA-containing glycerophospholipids and eicosanoids such as prostaglandin A2 (PGA₂) was observed in the intestinal tissue of NEC pigs. Glycerophospholipids are major components of cell membrane lipids that help maintain structural integrity of biological membranes as well as provide precursors for eicosanoid synthesis (52). LCPUFAs such as AA are mainly esterified in the sn-2 position of glycerophospholipids and circulating levels of AA are tightly controlled by cleavage of AA at the sn-2 position by phospholipase (PLA₂) or reacylation of fatty acids back into phospholipids (53–55). In NEC, it appears that the high rate of reacylation into phospholipids, as evidenced by increased AA-containing glycerophospholipids, is not sufficient to balance circulating levels of AA. This imbalance results in increased downstream metabolites of AA such as adrenic acid (22:4n-6), a 2-carbon elongation product of AA and a precursor of prostaglandins and thromboxanes (56). Readily available AA and its downstream eicosanoids may be useful in the initial phase of injury or inflammatory response; however, prolonged AA metabolites without a concomitant increase in proresolving mediators may confer a persistent proinflammatory state that could exacerbate tissue injury as seen in NEC (57).

A limitation of this study is the low incidence of NEC in this model which may be due to a delayed (starting on day 3) and slow advancement of enteral feedings to 50% of total energy needs by the final day of the protocol. The EN was limited to ensure adequate time of parenteral lipid emulsions to induce tissue metabolomic changes. The results of this 8-day study suggests that short exposure durations of parenteral lipid emulsions are unlikely to mediate a change in risk of NEC. The impact of receiving parenteral lipid emulsions for prolonged periods greater than 8 days on the risk of NEC remains unknown.

In summary, we show that the composition of parenteral lipid emulsion significantly influences the RBC membrane and intestinal tissue fatty acid levels and metabolomic profiles but not the intestinal NEC incidence or histology. Our results also show that development of NEC in the preterm piglet model is accompanied by alterations in proteomic and metabolomic profiles. Specific biological pathways identified that may present novel mechanistic insights into the pathogenesis of NEC in preterm infants include decreased AMPK signaling, increased serotonin, and imbalance in arachidonic acid-containing glycerophospholipids. Targeted validation of these metabolic signatures is warranted to further investigate the role of these pathways in NEC pathogenesis.

GRANTS

This research was funded by Charles H and Judy Hood Family Infant Health Research Program, National Institute of Diabetes and Digestive and Kidney Diseases (NIH R01 DK104346), and in part by federal funds from the US Department of Agriculture, Agricultural Research Service under Cooperative Agreement Number 3092-51000-060-01.

DISCLOSURES

Fresenius Kabi provided the lipid emulsions used in this study. C. R. Martin has grant support from Feihe International and Mead Johnson Nutrition; serves on the scientific advisory boards of Plakous Therapeutics, Inc. and LUCA Biologics; and has served as a consultant to Fresenius Kabi. D. Burrin has received grant support from Fresenius Kabi outside of this work.

AUTHOR CONTRIBUTIONS

D.B., S.D.F., and C.R.M. conceived and designed research; P.S., J.B., B.S., D.B., M.H.P., X.G., S.T.D., T.A.L., and C.R.M. performed experiments; W.Y., P.S., J.B., B.S., D.B., M.H.P., H.H.O., S.T.D., T.A.L., S.D.F., and C.R.M., analyzed data; W.Y., P.S., J.B., B.S., D.B., M.H.P., H.H.O., S.T.D., T.A.L., S.D.F., and C.R.M. interpreted results of experiments; W.Y., P.S., B.S., and C.R.M. prepared figures; W.Y., P.S., S.D.F., and C.R.M. drafted manuscript; W.Y., P.S., J.B., B.S., D.B., M.H.P., H.H.O., S.T.D., T.A.L., S.D.F., and C.R.M. edited and revised manuscript; W.Y., P.S., J.B., B.S., D.B., M.H.P., H.H.O., S.T.D., T.A.L., S.D.F., and C.R.M. approved final version of manuscript.