Fatty acid composition of adipose tissue at term indicates deficiency of arachidonic and docosahexaenoic acid and excessive linoleic acid supply in preterm infants

Three pre-defined groups of infants were studied: (Near-)TI with gestational age (GA) at birth ≥ 34 ± 0/7 weeks, with clinically indicated surgical intervention soon after birth (postnatal age (PNA) at surgery < 14 days). This GA range was required because most infants with abdominal wall defects or hydrocephalus are delivered before term. The postnatal age at surgery was chosen because surgical intervention is predominantly, yet not always, in the first week of life (Table 1). PTI with GA at birth of 23+0/7 to 31+6/7 weeks, without gastrointestinal problems, and with clinically indicated surgery (e.g., herniotomy) at term-equivalent age (± 6 weeks, i.e., at a postmenstrual age of ≥ 34 ± 0/7 weeks, and a postnatal age of > 14 days). Preterm infants with enterostomy (PTI/E) with GA at birth 23+0/7 to 31+6/7 weeks with enterostomy because of necrotizing enterocolitis and/or intestinal perforation and enterostomy closure at term-equivalent age ± 6 weeks. As defined a priori, these PTI/Es were considered separately because we anticipated prolonged periods of parenteral fat supply from SMOFlipid^® (Fresenius Kabi, Bad Homburg, Germany) and Omegaven^® (Fresenius Kabi) as well as potential malabsorption/intestinal losses of related nutrients through the enterostomy before adipose tissue sampling, potentially confounding lipid homeostasis. While it was initially intended also to collect adipose tissue and blood samples in the PTI/E group during the initial laparotomy for necrotizing enterocolitis and/or intestinal perforation, this turned out to be impractical, because the team felt that asking for study consent was inappropriate in the respective emergency situation. Consequently, all three groups were sampled only once at term-equivalent age (± 6 weeks).

Postnatal nutrition occurred according to local standards as previously described [28]. Infants were fed expressed breast milk and/or formula starting at day one. Enteral feeding was increased by 20–30 ml/kg/d, as tolerated, and breast milk was supplemented with commercial fortifiers to meet recommended intakes [29]. Consequently, partial parenteral nutrition (Amino acids, glucose, electrolytes and fat emulsion) starting at day 1 and complementing enteral nutrition until full feeds were reached was usually discontinued at days 5–7 (only the PTI/E infants received prolonged parenteral nutrition).

Outcomes were the molecular compositions of fatty acids in TG and PC in mol% in adipose tissue (primary outcome) and in plasma and erythrocytes (secondary outcomes).

Adipose tissue, plasma and erythrocyte membranes were extracted with chloroform:methanol according to Bligh and Dyer as previously described [24, 30, 31]. Erythrocyte membranes were prepared by suspending 100-µL erythrocyte pellet in 5-mL double-distilled, degassed water (4 °C). Samples were homogenized, kept for 15 min at 4 °C, and then centrifuged at 3000 × g for 20 min. The supernatant was decanted and the pellet used for extraction.

Phospholipid molecular species were analyzed by LC–H-ESI-MS/MS, using PC20:0/20:0 and PE14:0/14:0 as internal standards. The equipment comprised a Finnigan Surveyor Autosampler plus MS Pump plus TSQ Quantum Discovery MAX equipped with heated electrospray ionization interface (H-ESI)) (Thermo Fisher Scientific, Dreieich, Germany), as previously described [32]. PC, lyso-PC and SPH were separated from other phospholipids with a Polaris 3 Si-A column (2.0 × 100 mm; Agilent Technologies, Böblingen, Germany) at 40 °C, using a mobile phase consisting of chloroform:methanol:300-mM ammonium acetate (60:38:2, vol/vol) at flow rates of 400 µL/min (0–0.05 min) → 600 µL/min(0.06–1.39 min) and 400 µL/min (1.40–7.00 min). Individual molecular species were analyzed at positive ionization in the selected reaction monitoring (SRM) mode, using phosphorylcholine (mass/charge [m/z] = + 184) as the diagnostic fragment for individual PC molecular species. Concentrations were calculated relative to their internal standards, and data corrected for the ¹³C effect of components differing in number of carbon units, and for differences in ionization rate according to chain length as described before [24]. PC molecular species were grouped together (see online supplement, Table e1), depending on the presence of two saturated fatty acids (sat-PC), or a mono-unsaturated oleic (OA/C18:1-PC/PE), di-unsaturated linoleic (LA/C18:2-PC), arachidonic (ARA/C20:4-PC) or docosahexaenoic acid (DHA/C22:6-PC) residue, which accounted for > 95% of whole PC.

Reproducibility of analyses was assessed by external standard (control plasma), and 95% confidence intervals for PC species and sub-group composition and concentration are provided in the online supplement (Table e1).

Neutral lipids were separated from phospholipids using the chloroform phase of Bligh & Dyer extracts and disposable 100-mg NH2-propyl cartridges (Strata^®, Phenomenex, Aschaffenburg, Germany) as described before. Fatty acids were determined as fatty acid methyl esters (FAME) after transmethylation using docosatrienoic acid (C22:3n−3) as previously described [33], and 1 µL injected into a HP5890 gas chromatograph (GC) (Hewlett Packard) equipped with a flame ionization detector (250 °C), hydrogen/synthetic air as combustion gases, and a 100 m × 0.25 mm × 0.20 µm HP-88 column (Agilent, Germany) with helium as carrier gas. Column settings were 130 °C for 3 min, and then linearly increased to 176 °C at 27 min, 186 °C at 37 min, 190 °C at 58 min and 220 °C at 92 min. This allowed for the separation of the fatty acids indicated in the results. Quantification was performed using calibration curves for the respective fatty acid methyl esters. Reproducibility of fatty acid quantification is provided in the online supplement (Table e2).

Sample size calculation was done assuming that the proportions of ARA and DHA in adipose tissue PC reflect the previously reported postnatal changes in plasma PC in preterm infants [24], but with slightly lower values (i.e. ~ 90%). Hence, we anticipated a proportion of DHA in adipose tissue TG in TI of 10 ± 2%, in PTI of 5 ± 2%, ARA-TG in TI of 30 ± 5% and in PTI of 20 ± 5%. The primary hypothesis was (DHA/ARA in TI) ≠ (DHA/ARA in PTI = DHA/ARA in PTI/E). Assuming normal distribution of the data, seven patients were required per group for an analysis of variance with post hoc test and partially connected groups to test the primary hypothesis for DHA and ARA (power = 90%, alpha = 2.5%). The level of significance alpha 2.5% results in a global level of significance of 5% when two tests are done (one for DHA and one for ARA) in the same cohort.

Because only 3 patients were recruited into the PTI/E group, PTI/E infants were analyzed descriptively only. Because proportions of DHA and ARA were not normally distributed (according to Shapiro–Wilk test), between-group comparisons (restricted to TI versus PTI) were performed by Wilcoxon test. Data are shown as median [25th percentile–75th percentile]. JMP (Version 13.0.0, SAS Institute GmbH, Germany) was used for statistical analysis. The correlation of fatty acid concentrations and postnatal age was analyzed visually. Graphs and tables were created with Excel 2007, Word 2007 (Microsoft Corporation, USA) and JMP 13.0.0.

PTI, compared to TI, had less ARA in TG of adipose tissue (p = 0.0003). This difference also applied to adipose tissue PC (p = 0.03). Similarly, the proportions of ARA-PC in plasma (p = 0.011) and erythrocyte membranes (p = 0.0007) were lower in PTI than in TI (Fig. 1, panels a–c). The fraction of ARA in plasma TG, however, did not differ between groups (p = 0.073). The proportion of ARA decreased with postnatal age in adipose tissue TG, plasma PC and erythrocyte PC (Fig. 1, panels d–f).

Similar differences were observed for DHA. Here, its proportion in adipose tissue TG of PTI was lower than that of TI (p = 0.006). Similarly, in PTI compared to TI, DHA-PC of adipose tissue (p = 0.0005), plasma (p = 0.002) and erythrocyte membranes (p = 0.0002) and the DHA fraction of plasma TG (p = 0.001) were lower (Fig. 2, panels a–c). As for ARA, the proportion of DHA in adipose tissue TG, and that of DHA-PC in plasma and erythrocytes decreased with postnatal age (Fig. 2, panels d–f).

By contrast, in lipids of PTI, the proportion of LA was generally higher than in TI. This applied to TG (p < 0.0001) and PC (p = 0.0002) of adipose tissue as well as to PC of plasma (p = 0.0002) and erythrocyte membranes (p = 0.0003). LA in plasma TG was not different between groups (p = 0.083) (Fig. 3, panels a–c). The proportion of LA increased with postnatal age in adipose tissue TG and in PC of plasma and erythrocytes (Fig. 3, panels d–f).

The result of the PTI/E group (n = 3) is shown in Figs. 1, 2, 3a–c) (and in the online supplement Fig. e1–6), with higher fractions of DHA in adipose tissue TG as well as PC of plasma and erythrocyte membranes than in PTI and TI. However, proportions of ARA were decreased and those of LA similar to PTI values.

To compare the proportions of DHA, ARA and LA in different compartments, data are demonstrated in Fig. 4. For ARA and DHA, the proportions only correlate between plasma PC and erythrocyte PC, but not between plasma PC and adipose tissue TG as well as between erythrocyte PC and adipose tissue TG. For LA, all proportions correlate well between all three compartments. The concentrations of DHA, ARA and LA in different compartments and groups are summarized in Table 2.

Correlations of postnatal age with the fractions of the major TG fatty acids, palmitic (C16:0), oleic (C18:1−N9) and linoleic (C18:2−N6), and of ARA (C20:4−N6) and DHA (C22:6−N3) are shown in the online supplement, Fig. e1 and e2. While in TG of adipose tissue, the fractions and concentrations of C18:1−N9 and C18:2−N6 increased with increasing postnatal age (Fig. e1A + C), fractions of C16:0, ARA and DHA decrease (Fig. e1A + B). However, absolute concentrations of ARA and DHA remained constant in adipose tissue (Fig. e1D). Such correlations only in part applied to TG of plasma (online supplement Fig. e2).

The proportion of LA increased with postnatal age. This applied for individual PC molecular species, and was strongest for those molecular species containing a palmitic or stearic acid rather than an oleic acid in combination with linoleic acid (see online supplement, Fig. e3A–e6A).

In this prospective observational study, the primary goal was to examine the fatty acid composition of adipose tissue, erythrocyte membranes and plasma in preterm infants in comparison to term infants. Unfortunately, only 3 preterm infants with enterostomy could be enrolled (because so few infants required enterostomies during the study period); hence, this group of infants was only analyzed descriptively. Furthermore, the PTI group is dominated by boys due to their higher incidence of inguinal hernia. Finally, it was impossible to measure body composition of the study infants at the time of surgery to calculate pool sizes of individual fatty acids (because these infants required monitoring, nasogastric tubes or respiratory support at that time). But relative molecular composition of fatty acids (in adipose tissue stores and plasma) may be more important for resulting composition of growing organs (like the brain) and better indicators of deterioration of the lipidome than absolute pool size. Strengths of this study are the prospective data and sample collection and handling, and that lipid analyses are based on established methods as demonstrated in previous studies [20, 24].