In the present study (Fig. 1), medical journals for all planned cesarean sections were in advance reviewed to define exclusion and inclusion criteria. Afterwards, third trimester pregnant women were recruited by an interview and blood samples prior to a planned cesarean section. After delivery, their offspring were enrolled by cord blood samples. The women were included at Obstetrics Dept., Naestved Hospital, Denmark from January – February 2014 (pilot study) and again from June 2014—July 2015 (main study period).

Inclusion criteria: maternal age of at least eighteen years, an expected healthy, singleton pregnancy verified by a routine ultrasound scan in the second trimester and maternal health without any diseases or medication causing adverse effects to the fetus.

Exclusion criteria: multiple pregnancies, medical treated thyroid disease, diabetes (including gestational), preeclampsia or any other disease suspected to cause adverse effect to the fetus. Lack of informed verbal and written consent from both parents led to exclusion, too.

In the study period, a total of 180 planned cesarean sections were performed. Sampling logistics and laboratory capacity (n = 31), challenged family situation and missing full parental consent (n = 14) as well as positive exclusion criteria (n = 39) were the cause that only 96 women were invited to participate. Of these, 68 accepted (70.83%). In the pilot study, 9 were included, obtaining a total of 77 participating women. According to guidelines at inclusion time (TSH > 3.0 mIU/L) [16], 24 women fulfilled the criteria for SCH, and 13 according to recent Danish guidelines (TSH > 3.7 mIU/L) [19, 24]. The rest were euthyroid. According to the different guidelines, study results were analyzed as cohorts SCH1 (subclinical hypothyroid group 1) and EU1 (euthyroid group 1) for TSH cutoff 3.0 mIU/L and SCH2 (subclinical hypothyroid group 2) and EU2 (euthyroid group 2) for TSH cutoff 3.7 mIU/L.

A 3.5 mL heparinized blood sample was collected for same-day analysis of thyroid status, glucose, cholesterol and triglycerides. A serum gel tube (3.5 mL) was collected for thyroid peroxidase antibodies (anti-TPO) measurements.

Thyrotropin (TSH), free triiodthyronin (fT3) and free thyroxine (fT4) were analyzed by Siemens Dimension Vista System by an electrochemical luminescent immunoassay based on LOCI-technology (2008 Siemens Healthcare Diagnostics). Third trimester normal values for maternal TSH was defined as TSH: 0.3—3.0 mIU/L (EU1) or 0.3—3.7 mIU/L (EU2). The lowest value followed the standard used at the Dept. of Biochemistry at Naestved Hospital, and the upper value was in accordance with the different guidelines [16, 24]. Maternal normal values for fT3 and fT4 followed the references used at Naestved Hospital: fT4 = 8.5 – 26.0 pmol/L and fT3 = 2.7—6.1 pmol/L. Anti-TPO was measured on Kryptor by TRACE-technology (2005 Brahms Kryptor) with a detection limit of 11 kU/l. An anti-TPO > 60 kU/l was considered a positive test. To compare metabolism, glucose, total-cholesterol, low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL) and triglycerides were analyzed by Siemens Dimension Vista System (2008 Siemens Healthcare Diagnostics) by photometry.

For reverse transcription by quantitative Polymerase Chain Reaction (qPCR), a 3 mL Tempus™ Blood RNA Tube (Life Technologies, Denmark) containing a reagent that instantly stabilizes intracellular RNA, was used. Samples were subsequently frozen and stored at -80 °C until analysis. The laboratory technician was blinded to which thyroid groups the samples belonged. Sample collection and handling were performed according to the manufacturer´s recommendation.

RNA was purified using PerfectPure™ RNA Blood kit (5 Prime, AH Diagnostics) in accordance with the manufacturer’s recommendations. The integrity of RNA was characterized by the RNA integrity number (RIN) measured on an Agilent 2100 Bioanalyzer (Agilent Technologies, Copenhagen, Denmark). For analysis, a RIN value above 7.0 was considered acceptable. Whole blood gene expressions of PGC-1β, TFAM, NRF-2 and SOD2 were examined. Table 1 presents the primer sequences used.

Reverse transcription was performed using Quanti-Tect Reverse Transcription Kit (Qiagen, Copenhagen, Denmark). A SYBR Green based qRT-PCR of PGC-1β, TFAM, NRF-2 and SOD2 mRNA was measured on a LightCycler 480 (Roche). To correct for potential variation in cDNA loading and quantity, the measured gene transcripts levels were normalized to the expression of ribosomal 18 S RNA. The qRT-PCR reactions were performed according to the LightCycler standard protocol. The qPCR reaction mixture included 1 × LightCycler FastStart DNA Master PLUS SYBR Green I (Roche, Copenhagen, Denmark) and 1 μl mRNA specific- and 0.5 μl of the 18 S primers, 1 μmol/l respectively. For each reaction 5 μl template cDNA and sterile water were added in a total reaction volume of 20 μl. Cycling conditions were 95 °C for 10 min, followed by 45 cycles at 95 °C for 10 s, and 60 °C for 1 min. All quantitative RT-PCR measurements were performed in triplicates. Melting curves were completed for the control of unspecific DNA amplification after each run. Unit of measurement for gene expressions was arbitrary units (a.u.).

Some results did not reach acceptable quality, and were excluded. The excluded samples were from 2 women with TSH > 4 mIU/L, and from 2 with TSH < 2 mIU/L, BMI-range 25.1–27.6. Therefore, only 73 samples presented full gene profile results (EU1 n = 51, SCH1 n = 22, EU2 n = 62, SCH2 n = 11), and for TFAM one result could be added to euthyroid-groups (EU1 n = 52, EU2 n = 63).

The amount obtainable of cord blood were too low for qPCR analysis. This was due to a very sparse amount of blood left for the Tempus tube after collection of blood for standard pH-measurements (2–4 mL), for thyroid function-tests (3.5 mL), anti-TPO measurements (3.5 mL), for flow cytometry (4 mL) (results published in [26, 32]) and for miRNA analyses (4 mL). Seldom, blood was left in the cord after collection for these other tubes. If more blood was left, it was very difficult to do more sampling due to coagulation.

The choice of mitochondria-related genes were based on the fact that these different genes reflect mitochondrial function differently [5, 11, 33], and due to former laboratory experience [6, 34].

For miRNA-screening, only a minor number of participants were chosen due to cost of analyses. Twenty-two maternal samples (11 euthyroid, 11 subclinical hypothyroid) and the associated cord samples were chosen randomly, as they fulfilled quality criteria concerning sufficient blood volume and no signs of hemolysis by the manufacturer´s primary quality check.

One EDTA-plasma fraction was sampled for miRNA-analysis and subsequently frozen and stored at -80 °C until shipping for analysis.

The real-time PCR panel analysis was performed by Exiqon the following way: Each RNA sample was successfully reverse transcribed (RT) into cDNA and run on the miRCURY LNA™ Universal RT microRNA PCR Human Panel I + II (752 assays). For normalization of data, the average of the assays detected in all samples was applied as it was the most stable normalizer. Numbers of miRNA present were detected and the quantification cycle (Cq) value of the global mean for each of the samples was identified.

Each individual amplification product on PCR panels was scrutinized by melting curve analysis, calculation of amplification efficiency and comparison of Cq value to background level in a negative control sample. For quality control, two types of RNA-spike-ins kits (Exiqon) were used: For RNA isolation control, UniSp2, UniSp4, UniSp5 (Exiqon) were added to the purification to detect any differences in extraction efficiency. For cDNA synthesis control, UniSp6 (Exiqon) was added in the RT reaction, giving the opportunity to evaluate it. In addition, a DNA spike-in (UniSp3, Exiqon) was present on all panels, to indicate inhibitions at the qPCR level by deviations in this reaction.

Exiqon was blinded to which thyroid groups the samples belonged.

Two-tailed t-test was used to compare data with a normal distribution. Otherwise, results were handled non-parametrically (non-normality confirmed by Shapiro-Wilks test). Wilcoxon Rank-sum test was used to compare level of differences. Spearman´s rho ρ was used to access correlations. A p-value < 0.050 was considered statistically significant. Data processes were carried out using STATA version 15.0 for Windows (StataCorp LP, College station, TX, USA).

Regarding miRNA, Exiqon analysed the data as follows: A Principal Component Analysis (PCA) was used to reduce the dimension of the large data set and identify the miRNAs with the largest variation across samples. The most differentially expressed miRNAs were analysed by t-test and a Benjamini–Hochberg correction. The analyses were made in accordance with the definition of SCH at that time (TSH > 3.0 mIU/L).

A calculation with 90% strength at the 5% significance level (alpha) specified recruitment of 242 euthyroid and 48 subclinical hypothyroid women to show a significant decrease in subclinical hypothyroid group TFAM. Calculations were based on a mean value of 1.70 in euthyroid and 1.19 in subclinical hypothyroid group, and a common SD of 1.00. For PGC-1β, NFR-2 and SOD2, recruitment of 252 euthyroid and 50 subclinical hypothyroid women was necessary to demonstrate a significant difference. Calculations were based on a mean value of 4.70 in the euthyroid and 3.80 in the subclinical hypothyroid group, and a common SD of 2.00. Calculations were based on a rate of subclinical hypothyroidism /euthyroidism = 0.20 (TSH cutoff 3.7 mU/L).

Calculations were performed retrospectively, as this study was performed in supplement to a flow cytometric evaluation of mitochondrial function (primary outcome) of the same cohort (unpublished results [26]), preceded by a method study [32]. Unfortunately, it was not possible to obtain the calculated power of the study, due to health circumstances in the research team, which interrupted the main study prematurely.