Population, study design and setting
This was a case-control study conducted in three hospitals, University of Naples “Federico II”, University of Salerno and Hospital “G. Moscati” of Avellino, in Campania, Southern Italy, from January 2011 to December 2013.
Pregnant women with a diagnosis of fetal malformations or fetal chromosomal abnormalities (Cases) were compared with healthy controls (CTRLs).
All Cases were recruited at the time of second trimester termination of pregnancy, while CTRLs delivered normally developed fetuses, and were recruited during their second trimester routine anomaly scan.
The presence or absence of fetal malformations or chromosomal abnormalities was defined based on ultrasound examination or karyotype and confirmed by postmortem autopsy by an expert pathologist or after pediatric examination of the newborns.
Exclusion criteria for Cases were: maternal age > 40 years old, twin pregnancy, pregnant women committed to carrying the pregnancy to term, TORCH (Toxoplasma, Rosolia, Citomegalovirus, Herpes) complex infection, or CNS defects with a known genetic cause.
The study was approved by the local ethics committee (IRB n.4/2013). A written consent form was signed by each participant at the time of enrollment.
Enrolled patients completed a questionnaire addressing clinical history and demographic characteristics and a complete obstetric visit was performed at enrollment to collect a thorough medical history. The investigations determined the presence of any known etiological factors of malformations, including: history of infections; malnutrition or metabolic disease (e.g. diabetes); drug use (e.g., thalidomide, anticoagulants, chemotherapeutic agents) and drug addiction (e.g., cannabis, cocaine, heroin); radiological investigations (e.g. X-rays, Computerized Tomography); and/or history of genetic syndromes.
Cases were subsequently divided into two groups:
-
CNS group which included all CNS malformations with unknown etiology;
-
Other group, which included all other malformations or chromosomal abnormalities.
Samples collection
Human tissue collection strictly adhered to the guidelines outlined in the Declaration of Helsinki IV edition. All blood was collected using a BD Vacutainer (Becton Dickinson, Oxfordshire, UK) blood collection red tube (no additives). Blood samples of the Cases were collected in the second trimester immediately before termination of pregnancy and before any drug administration. Blood samples of CTRLs were collected during the second trimester routine anomaly scan. After centrifugation, serum was transferred to cryovials and immediately frozen to − 80 °C until the time of analysis. All patients were asked to respect a 12-h fast before blood collection.
Metals concentration
Serum samples were analyzed with an ICP-QMS Bruker 820-MS (Bruker Daltonics, Billerica, MA). Operational parameters were: Plasma flow: 18 L/min, Auxiliary flow: 1.8 L/min, Sheath Gas: 0.14 L/min, Nebulizer flow: 0.98 L/min, RF power: 1.40 kW, Pump rate: 4 rpm, Stabilization delay: 20 s, First Extraction Lens: − 40 V, Second Extraction Lens: − 166 V, Third Extraction Lens: − 234 V, Corner Lens: − 208 V, Mirror Lens left: 29 V, Mirror Lens right: 26 V, Mirror Lens bottom: 30 V; CRI parameters: Skimmer Gas: H2 at 50 ml/min, Sample Gas: He at 10 ml/min; dwell time, 50,000 μs; no. of scan replicate: 10, no. replicate for sample: 5. High purity He (99.9999% He, AirLiquide Srl, Italy) and H2 (99.9999% H2, produced by the DBS H2 generator PGH2–300) were used, in order to minimize the potential problems caused by unidentified reactive contaminant species in the cell. The high radio frequency power (1400 W) helped maintain plasma stability. All chemicals were of the highest purity grade that is commercially available. Before use, all glassware and plastic containers were cleaned by washing with 10% ultra-pure grade HNO3 for at least 24 h, and then rinsed copiously with 18 MΩ water (produced by the Direct-Q-UV, Millipore, Billerica, MA, USA) before use. Peltier cooled ICP spray chamber temperature was setted at 3 °C. This ensure temperature stability and to reduce the water vapor production in the nebulizer gas flow. Standard solutions and samples were analyzed by means the SPS3 autosampler (Varian Inc., Mulgrave, Australia) coupled to the ICP mass spectrometer After collection, all serum samples were anonymized (a 3-letter and 3-number code was assigned) and stored at − 80 °C until analysis. Serum samples were pre-heated before the analysis keeping them at room temperature for two hours before sample preparation. After a gently vortex mixing (30 s at 300 rpm), 500 μL was diluted with 100 μL 0.1% (V/V) Triton-X-100 solution (Sigma-Aldrich, Seelze, Germany). and filled up to 5 mL with a 0.5% (v/v) NH4OH (Merck, Darmstadt, Germany) in a 10 mL polypropylene tube using a 5 mL bottle-top dispenser (Brand, Wertheim, Germany). Samples were then homogenized with an oribital stirrer KS3000i (IKA, Staufen, Germany).
Standard addition method was used for calibration according to Heitland and Köster [17]. Briefly, five hundred microliters of the serum sample were diluted up to 5 mL. A stock standard solution containing the metals under investigation were prepared in 100 mL polypropylene flasks diluting 20 mg/L multi-element standard solutions for Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) (Ultrascientific, North Kingstown, USA). These solutions were at defined volumes to a mixture composed of of 500 μL serum, 100 μL 0.1% (v/v) Triton-X-100 solution in 10 mL plastic (polypropylene) tubes for autosampler (Merck, Darmstadt, Germany). These solutions were finally filled up to 5 mL using 0.5% NH4OH. Dilutions were carried out with variable volume pipettes (volumes 50–1000 μL) and a bottle dispenser with adjustable volumes from 1 to 5 mL (Eppendorf, Hamburg, Germany). The isotopes analyzed were: 7Li, 9Be,27Al, 49Ti, 51V, 52Cr, 55Mn,59Co, 65Cu, 66Zn, 71Ga, 78Se, 85Rb, 88Sr, 90Zr, 93Nb, 98Mo, 101Ru, 107Ag, 114Cd, 115In, 121Sb 125Te, 137Ba, 140Ce, 141Pr, 146Nd, 147Sm, 153Eu, 157Gd, 163Dy, 166Er, 169Tm, 172Yb, 178Hf, 181Ta, 182W, 185Re, 189Os, 193Ir, 195Pt, 202Hg, 205Tl, 206,207,208Pb. These 15: 9Be, 27Al, 51V, 52Cr, 55Mn,59Co, 65Cu, 66Zn, 78Se, 107Ag, 114Cd, 121Sb,202Hg, 205Tl, 206,207,208Pb were quantified by calibration curve while the others were quantified by semi-quantitative method which uses an average response factor based on neighboring elements on the periodic table. This approach can have the same accuracy as calibration with individual standards [18]. Samples were analyzed in a random computer-generated sequence. Another multi-element calibration solution of a different production lot from the same manufacturer was used to verify the elemental concentrations in all multi-element calibration solutions. All solutions were analyzed by aspirating (with a Y–connection) an internal standard solution of 10 μg/L of 6Li, 45Sc, 72Ge, 89Y, 103Rh 159Tb, 165Ho, 209Bi in 2% (v/v) HNO3 (Ultrascientific, North Kingstown, USA). Internal quality assurance was obtained analyzing the Clinchek® Whole Blood Control Level 1–3 (Recipe, Munich, Germany) and Seronorm® Trace Elements Whole Blood Control Level 1–3 (Sero AS, Billingstad, Norway).. Repeat analysis of method blanks showed that all materials and reagents were free of metal contamination. The limit of detection (LOD) for each element was evaluated with two methods: first, as the concentration that corresponds to a signal equal to the average blank signal plus 3 times the standard deviation of the signal from 10 replicates of a blank sample (LOD = Avgblank + 3σblank); second as 3 times the concentration relative to the signal from the standard deviation of the y-intercept of the calibration curve (sx/y) [19]. The higher value was used as LOD. All analyses were replicated three times and mean values calculated to be used for statistical testing.
Statistical analysis
Study data were collected and managed using REDCap electronic data capture tools hosted at the INFN Institute of the University of Salerno (Italy) [20]. Statistical analysis was performed using Statistica software (StatSoft, Oklahoma, USA) and Minitab (Minitab Inc., Pennsylvania, USA).
The normal distribution of all data was verified using the Kolmogorov-Smirnov test. Logarithmic transformation was performed on skewed variables. For descriptive purposes, continuous variables were reported as mean ± standard deviation (SD) as untransformed values, while categorical variables were reported as number (percentage). Independent two-tailed t-tests were used to compare parameters between two groups. A ɑ-value< 0.05 was considered statistically significant. The ɑ-value was adjusted according to Bonferroni setting to 0.05/88 = 0.0006. Values for concentrations below the limit of detection (LOD) were imputed as LOD divided by square root of 2 [21]. For parameters of more than one category, Analysis of Variance (ANOVA) was performed. Significant differences were followed by a Tukey’s post-hoc.
Multivariate statistics analysis
Two classification models were used to verify if metal distribution was globally different among the CTRLs, Cases (CNS malformations) and Other classes. One model uses Principal Component Analysis (PCA), an unsupervised algorithm finding the directions that best explain variance in a dataset (X) without referring to class labels (Y). The dataset, comprised of all elemental concentrations of all samples, was median-centered and log transformed. The data were then summarized into many fewer variables (called principal components) which are weighted averages of the original variables (where the weighting profiles are called loadings). The PCA analysis was performed using the prcomp package for R Software (Foundation for Statistical Computing, Wien, Austria). Calculations were based on singular value decomposition. A separate model used to determine if metal distributions are globally different between the Control, Case and Other classes was Partial Least Squares-Discriminant Analysis (PLS-DA), which is a supervised method that uses multivariate regression techniques to extract, via linear combination of original variables (X, the metal concentrations), the information that can predict class membership (Y, CNS malformation or Control). We used the R pls package to perform the PLS regression, using the plsr function [22].. The wrapper function offered by the caret package was used to perform classification and cross-validation [23]. A permutation test was done to evaluate the significance of class discrimination. We built, for every permutation, a PLS-DA model between the data (X) and the permuted class labels (Y), using the optimal number of components determined by cross validation for the model based on the original class assignment. Two types of test statistics for measuring class discrimination were used. The first one is based on prediction accuracy during training. The second one is separation distance based on the ratio of the between-group sum of squares and the within-group sum of squares (B/W-ratio). Variable Importance in Projection (VIP) scores were calculated for each component. The definition of VIP score is that of a weighted sum of squares of the PLS loadings, derived from the explained Y-variation in any dimension. Also, we take into accout the weighted sum of PLS-regressions, whoseweights are defined as the functions of the sum reductions of squares through the numbers of PLS components. The average of the feature coefficients (i.e. loadings) was used to indicate the overall coefficient-based importance.