Untargeted metabolic screening of preterm infants on a human milk diet
Aim. This dataset includes metabolic fingerprints of preterm infants exclusively receiving (≥ 80% of total intake) either donor human milk (DHM) provided by a Human Milk Bank (N=20) or own mother’s milk (OMM) (N=20) recorded using a LC-QTOFMS instrument from urine samples collected at one month of age. The dataset was used to assess nutrition-derived changes in preterm infants at the phenotype level in a non-invasive manner. Study population and urine collection. A prospective, observational cohort study was conducted including consecutively admitted preterm infants born at ≤32 weeks of gestation and/or birth weight ≤1500 g in the Division of Neonatology of the University and Polytechnic Hospital La Fe (Valencia, Spain). Two groups were recruited according to the main feeding type accounting for ≥80% v/v of the nutritional intake with either OMM or pasteurized DHM after achieving full enteral nutrition (150 mL/kg/day). Urine samples were collected from infants exclusively receiving DHM (N=20) or OMM (N=20) one month after birth using sterile cotton pads placed in the diaper. Cotton pads were retrieved after one hour and squeezed with a syringe. The process was repeated until a minimum volume of 1 mL was collected. Urine samples were stored at −80 ⁰C until analysis. Sample preparation. Urine samples were thawed on ice and thoroughly shaken on a Vortex® mixer during 10 s followed by centrifugation at 16 000 x g and 4 ⁰C during 10 min. 50 μL of supernatant were added to 50 μL of an internal standard (IS) solution containing reserpine, phenylalanine-D5, leucine-enkephalin, caffeine-D9, and methionine-D3 at 2 μM each in H2O:CH3CN (96:4, 0.1% HCOOH v/v) and transferred to a 96-well plate. A blank extract was prepared using water instead of urine and following the same procedure as described for urine samples. A pooled quality control (QC) sample was prepared by mixing 5 μL of each study sample.
Steps to reproduce
Metabolic screening. Metabolomic analysis was carried out employing a 1290 Infinity UPLC system from Agilent Technologies (CA, USA) equipped with a UPLC BEH C18 column (100 x 2.1 mm, 1.7 µm) from Waters (Wexford, Ireland). The flow rate was set to 400 µL min-1 running a binary gradient with 0.1% v/v HCOOH in H2O and 0.1% v/v HCOOH in CH3CN as mobile phase components. Column and autosampler were kept at 55 and 4 ⁰C, respectively and the injection volume was 4 µL. For detection, a QTOF-MS system working in the ESI+ mode was used in the range between 70 and 1700 m/z with automatic MS spectra recalibration during analysis. LC-TOFMS data acquisition was carried out employing MassHunter Workstation (version B.07.00) from Agilent. Data pre-processing. Centroid LC-TOFMS raw data were converted to mzXML format employing ProteoWizard. The selection of parameters for peak table extraction and alignment was based on the observed variation of retention times and m/z values of ISs. XCMS software and CAMERA in R 3.6.1 were employed for the generation of peak tables. For data acquired from urine samples, the centWave method with the following settings was used for peak detection: m/z range = 70-1200, ppm = 15, peakwidth = (3 and 20), snthr = 6. A minimum difference in m/z of 0.01 Da was selected for overlapping peaks. Intensity-weighted m/z values of each feature were calculated using the wMean function. Peak limits used for integration were found through descent on the Mexican hat filtered data. Peak grouping was carried out using the “density” method using mzwid = 0.015 and bw = 6. Retention time correction was carried out using the “obiwarp” method. After peak grouping, the fillPeaks method with the default parameters was applied to fill missing peak data. Automatic integration was assessed by comparison to manual integration using IS signals. Further data processing and statistical analysis were carried out in MATLAB 2017b (Mathworks Inc., Natick, MA, USA). During data pre-processing and filtering of data from urine analysis, features with an intensity < 800 AU and those with a mean peak area <9 x mean peak area in blank samples, were removed. Intra-batch effect correction was performed using Quality Control-Support Vector Regression algorithm and the LIBSVM library with the following parameters: ε–range = 2 to 5%; γ-range = 1 to 105; C interval = 50%. Finally, features with a %RSD in QC samples >20% after QC-SVRC were removed from the peak table. A total of 10450 features were retained for between group comparison. Urinary metabolic data were normalized to the creatinine concentration in each urine sample quantified by the modified Jaffe method (DetectX® urinary creatinine detection kit, Arbor Assays, Ann Arbor, MI, USA) following the manufacturer’s protocol and employing a dilution factor of 1:4 during sample preparation.