MicroRNA-148b-3p and MicroRNA-25-3p Are Overexpressed in Fetuses with Late-Onset Fetal Growth Restriction

Objective: It was the aim of this study to describe a microRNA (miRNA) profile characteristic of late-onset fetal growth restriction (FGR) and to investigate the pathways involved in their biochemical action. Methods: In this prospective study, 25 fetuses (16 normal and 9 with FGR [estimated fetal weight <10th centile plus cerebroplacental ratio <0.6765 multiples of the median]) were evaluated with Doppler ultrasound after 36 weeks. Afterwards, for every fetus, plasma from umbilical vein blood was collected at birth, miRNA was extracted, and full miRNA sequencing was performed. Subsequently, comparisons were done in order to obtain those miRNAs that were differentially expressed. Results: The FGR fetuses expressed upregulation of two miRNAs: miR-25-3p and, especially, miR-148b-3p, a miRNA directly involved in Schwann cell migration, neuronal plasticity, and energy metabolism (p = 0.0072, p = 0.0013). Conclusions: FGR fetuses express a different miRNA profile, which includes overexpression of miR-25-3p and miR-148b-3p. This information might improve our understanding of the pathophysiological processes involved in late-onset FGR. Future validation and feasibility studies will be required to propose miRNAs as a valid tool in the diagnosis and management of FGR.


Introduction
Fetal growth restriction (FGR) occurs when a fetus fails to reach its growth potential [1]. Its importance lies in its association with a higher probability of perinatal morbidity and mortality and the subsequent long-term neurologic and cardiovascular consequences in adult life [2,3]. FGR comprises two varieties: early-onset FGR (< 34 weeks) is less frequent and is characterized by the existence of placental disease, deceleration of fetal growth, and progressive hemodynamic dysfunction, typically af-DOI: 10.1159/000507619 fecting in its onset the uterine and umbilical Doppler examination, while late-onset FGR (> 34 weeks) is more frequent and is defined by the unbalance between fetal demands and placental supply, resulting in the detection of a characteristic low cerebroplacental ratio (CPR) regardless of the estimated fetal weight (EFW) [4]. Late-onset FGR tends to be subtle. However, despite what might be thought, it is especially harmful, as it leads to frequently undiagnosed suboptimal arborization and brain underdevelopment [5,6].
Unfortunately, an adverse perinatal outcome (APO) in fetuses with late-onset FGR is difficult to predict. Clinical protocols may use the CPR or a combination of the CPR and EFW for its identification. However, this methodology has a poor accuracy and cannot be applied clinically yet [7][8][9][10]. Hopefully, this prediction could be theoretically improved using diverse biochemical markers, a search that has become of crucial importance.
MicroRNAs (miRNAs) are small RNA sequences, on average 22 nucleotides in length [11], with the ability to regulate gene expression in different organisms. Their action is mediated through the inhibition of translation or the promotion of mRNA degradation [12]. Their genes are encoded within the genome, suggesting that their transcription might be coordinated with the transcription of other genes. In summary, generation of the mature miRNA molecule involves the processing of a primary miRNA transcript in the nucleus to obtain the final product in the cell cytosol, a small single RNA strand which participates in a variety of cellular processes (development, proliferation, function, and differentiation) and in the pathogenesis of many human diseases [13]. miRNAs can target genes with relative specificity. To date, about 2,500 miRNA sequences are known in humans (miRBase v21) [14], and it was predicted that 30-80% of the human genes may be influenced by at least one miRNA [15,16]. Interestingly, recent studies have shown that miRNAs are also expressed in the placenta, suggesting a potential regulatory role in its development [17]. In addition, some miRNAs have been described to be hypoxia-regulated and associated with FGR [18].
The purpose of the current study was to define a mi-RNA profile characteristic of late-onset FGR, investigating the pathways involved in their biochemical action.

Patient Recruitment and Doppler Examination
This was a prospective study of 25 fetuses at the tertiary public maternity La Fe Hospital. These fetuses underwent an ultrasound examination between 36 and 40 weeks that included biometry and EFW calculation plus a Doppler evaluation of the umbilical artery (UA) and middle cerebral artery (MCA) pulsatility indices (PIs). The UA and MCA were examined using color and pulsed Doppler ultrasound according to earlier descriptions [19,20], and the CPR was calculated as the simple ratio between the MCA PI and the UA PI [21].
All babies were delivered within 15 days or less after the scan, and only the last examination per fetus was included in the analysis. In order to adjust for the effect of gestational age (GA), EFW, and birth weight (BW), the values were converted into local reference centiles [22] adjusted only for fetal gender. Also, CPR values were converted into multiples of the median (MoM), dividing each value by the 50th centile at each GA as previously described [19]. CPR medians (50th centile) were represented by the equation CPR 50th centile = -3.814786276 + 0.36363249 × GA -0.005646672 × GA 2 , where "GA" was the GA in weeks with decimals.
All Doppler examinations were performed by the first author (J.M.-R.), a teaching expert in obstetric ultrasound certified by the Spanish Society of Obstetrics and Gynecology, using General Electric Voluson ® (E8/E6/730) ultrasound machines (General Electric Healthcare, Spain) with 2-to 8-MHz convex probes during fetal quiescence, in the absence of fetal tachycardia, and keeping the insonation angle relative to the examined vessels as small as possible and always below 30°.
GA was determined according to the crown-rump length in the 1st trimester. Multiple pregnancies and those complicated by congenital fetal abnormalities or aneuploidies were excluded. Gestational characteristics including parity, number of gestations, and maternal ethnicity, age, weight and height were collected at examination, together with the indicated ultrasound parameters. Labor outcome data including BW, BW centile, mode of delivery, 5-min Apgar score, cord arterial pH, and date of admission to the neonatal care unit were also collected at birth.

Ponderal and Hemodynamic Characteristics of the Groups Studied
For the purpose of comparison, the study included two different types of fetus: late-onset FGR fetuses, with an abnormal EFW (< 10th centile) and an abnormal CPR (< 0.6765 MoM), and normal fetuses, with a normal EFW (> 10th centile) and a normal CPR (> 0.6765 MoM) [19]. Fetuses with intermediate features (abnormal CPR with normal EFW or normal CPR with abnormal EFW) were not considered.

Sample Collection and Small RNA Extraction and Quantification
After birth, plasma samples from the fetal umbilical vein and maternal peripheral blood were collected in EDTA tubes and centrifuged at 3,500 rpm for 10-15 min. Once plasma had been obtained, each sample was stored at -80 ° C until small RNA extraction. While maternal plasma was stored for future research, 500 μL of fetal blood plasma were used to isolate cell-free total RNA (including miRNAs) using the miRNeasy Serum/Plasma kit (Qiagen, Valencia, CA, USA) following the manufacturer's protocol. The RNA was eluted with 25 μL of RNase-free water. The concentration of cell-free total RNA (including miRNAs) was quantified us-

Library Preparation and Next-Generation Sequencing
Small RNA libraries were generated and indexed using a modified Illumina TruSeq small RNA protocol. In this modified protocol, the libraries were size selected (range 90-170 bp) using a Blue Pippin instrument (Sage Science, Beverly, MA, USA). A positive RNA control was included (Thermo Fisher Scientific Human Brain Total RNA catalog #AM7962). Single-end sequencing was performed on an Illumina NextSeq platform on High-Output 1 × 50 bp Run (NextSeq 500/550 High-Output v2 75 cycles kit, FC-404-2005).

Differential Expression Analysis
The first step was to assess the quality of the Illumina raw sequences with the FastQC software (https://www.bioinformatics. babraham.ac.uk/projects/fastqc/). Based on the results obtained, the sequence reads were trimmed to remove sequencing adapters and low-quality bases using the software Cutadapt (http://cutadapt.readthedocs.org/en/stable/). Once the data were deemed of sufficient quality, they were mapped against the human GRCh38 build reference sequence, taken from Ensembl. After that, an intersection between the aligned position of reads and the miRNA coordinates taken from miRBase v21 was performed. The alignment and quantification steps were performed using the Subread and Rsubread packages [23,24].
A multidimensional scaling plot was used to get a closer look at how samples were distributed according to the miRNA expression values. miRNAs with very low counts across all libraries provide little evidence for differential expression. We filtered out these miRNAs prior to further analysis. Subsequently, the trimmed mean of M-values normalization method (TMM normalization) [25] was performed to eliminate composition biases between libraries. We also estimated the specific dispersions per gene with a negative binomial distribution [26,27].

Group Comparisons
After miRNA had been extracted and sequencing performed, comparisons were done between the group of late-onset FGR fetuses and the group of normal fetuses (control group). As previously indicated, late-onset FGR was considered when abnormal values of CPR [19,28] plus abnormal values of BW [22] were present. Conversely, normality was considered if both parameters were within normal range. As per the local protocol, all fetuses were subsequently managed according to their progression in labor, including intrapartum fetal heart rate assessment, which was interpreted according to the FIGO guidelines [29]. All study comparisons were performed using the Mann-Whitney test for continuous data and Fisher's exact test for categorical data. Significance was considered with p < 0.05.

Prediction of miRNA Targets and Overrepresentation Analysis
We first used DIANA-microT-CDS, accessed from the DIANA web server [30]. This tool showed whether the target was also predicted by miRanda or TargetScan or was experimentally validated in TarBase v7.0. We used the DIANA-miRPath v3.0 functional analysis online suite to identify miRNAs controlling significant molecular pathways annotated on the Kyoto Encyclopedia of Genes and Genomes (KEGG), using the following default param-eters: experimentally supported interactions from DIANA Tar-Base v.7.0; a p value threshold of 0.001; and a microT threshold of 0.8. To reduce the number of false-positive miRNA targets, we applied a false discovery rate (FDR) correction to selected KEGG pathways. The algorithm used in this analysis was a one-tailed Fisher exact test [31].

Descriptive Statistics of the Study Population
The study included 25 fetuses, of whom 14 (56%) were male and 11 (44%) were female, all of them being of Spanish Caucasian origin. Online supplementary Figure 1 (see www.karger.com/doi/10.1159/000507619 for all online suppl. material) shows the distribution of measurements according to the CPR MoM and BW centile [8,19,28]. Nine fetuses had late-onset FGR (abnormal CPR plus abnormal BW values), while 16 were fetuses with normal hemodynamic and ponderal features (normal CPR plus normal BW).
In Table 1 we compare the normal with the FGR pregnancies. In summary, the mothers in the FGR group were thinner than those in the normal group, and the fetuses had a lower CPR MoM, EFW, EFW centile, BW, and BW centile (p < 0.001). In addition, when compared with the fetuses with a normal outcome, the FGR fetuses were delivered earlier (p = 0.04). Table 2 shows the initial comparison between normal and FGR fetuses regarding the differential expression analysis of upregulated and downregulated miRNAs. For accuracy and selection purposes, only miRNAs with the criterion of FDR < 0.05 were included. In comparison with the normal fetuses, the FGR fetuses showed a total of four initially differentially expressed miRNAs: miR-148b-3p, miR-25-3p and miR-16-5p, which were upregulated, as well as miR-1910-5p, which was downregulated. miR-148b-3p had by far the highest significance. As a matter of caution, we discarded miR-16-5p due to its relation to hemolysis, which might always be present in blood samples to some degree [32,33]. Figure 1 shows the heat map of the miRNA expression profile with the miRNAs selected in Table 2. The cluster was done on the basis of log 2 (expression level in treatment/expression level in control). Yellow denotes downregulation of miRNAs and red denotes upregulation of miRNAs in blood samples from the neonatal cord.   Figure 2 shows box-and-whisker plots representing the logarithm of counts per million reads (logCPM) of those miRNAs that presented significant differences: miR-148b-3p and miR-25-3p (Mann-Whitney p < 0.05).

Comparison of the Logarithm of Counts per Million Reads
It is noteworthy to underline that a difference existed between the differential expression analysis of upregulated and downregulated miRNAs and the individual statistical analysis performed between normal and FGR fetuses concerning the logCPM. When the analysis of differential expression was done, four miRNAs were significant with Color key and histogram  FGR, fetal growth restriction; logFC, logarithm with the base of 2 of the fold change -a negative logFC value corresponds to a downregulated miRNA and a positive logFC value means that the miRNA is upregulated relative to the reference condition; logCPM, logarithm with the base of 2 of the counts per million reads obtained by the miRNA; F, the value of the statistic test; FDR, p corrected value with false discovery rate. DOI: 10.1159/000507619 an FDR < 0.05 (Table 2). However, when the logCPM comparison was performed in an independent way and without considering the expression of all those miRNAs that contributed to the differential expression analysis, only miR-148b-3p and miR-25-3p turned out to be finally statistically significant (Fig. 2). Therefore, only these were finally selected to explore relevant pathways related to FGR.

Analysis of miRNA Targets and Biochemistry Pathways in the Context of FGR
All miRNAs had a large number of potential target sites, so we explored those of relevance to FGR. Interestingly and in agreement with Figure 2, miR-1910-5p did not release any pathway. Therefore, miR-16-5p and miR-1910-5p where finally removed from the analysis. In order to clarify the role of the remaining miRNAs miR-148b-3p and miR-25-3p in FGR, we analyzed the biochemical networks in which they participate. We carried out a DIANA-miRPath v3.0 analysis and KEGG pathway analysis to look for any significantly enriched pathway. A total of 29 pathways with an FDR < 0.05 were retrieved ( Table 3). Some of them are related to lipid metabolism, such as biosynthesis of fatty acids [31] and sphingolipids [34], crucial for neuronal tissue development, while others are related to protein processing in the endoplasmic reticulum [35] or to protein metabolism, such as biosynthesis of branchedchain amino acids valine, leucine, and isoleucine [36]. Pathways obtained by miRPath v3.0 were ordered according to p value. Genes: number of genes from that pathway regulated by some of significant differentially expressed miRNAs in FGR vs. normal fetuses. miRNAs: number of differentially expressed miRNAs in FGR vs. normal fetuses implied in that pathway [30]. FGR, fetal growth restriction.

Principal Findings
Analyzing circulating miRNAs from neonatal cord blood and using next-generation sequencing we found that miR-148b-3p and miR-25-3p were upregulated in late-onset FGR fetuses. This different miRNA profile is a novel finding that might improve our understanding of late-onset growth compromise.
Research Implications CPR MoM has emerged as the best marker of APO at the end of pregnancy [37][38][39][40]; however, its accuracy as a single parameter or in association with EFW centile and other clinical parameters is not good enough to obtain clinically reliable results [8]. As a consequence, new biochemical markers are being investigated in order to obtain accurate predictions of APO and neurocognitive dysfunction.
In recent years, the focus has shifted to miRNAs. These molecules behave like fine tuners in the regulation of gene expression and are crucial for many biological processes [41]. A large number of the more than 1,000 miRNAs discovered in humans are related to pregnancy and are produced by the placenta and the uterus in normal and pathological conditions. They exert their action locally in a paracrine fashion, or distally released as exosomes in maternal blood, regulating fetal and maternal homeostasis [41]. Moreover, several studies have shown that altered expression of the miRNome in maternal circulation or in placental tissue may reflect gestational disorders such as preeclampsia, spontaneous abortion, preterm birth, low BW, or macrosomia [42].
Role in Neuronal Plasticity miR-148b seems to have special relevance in diverse molecular mechanisms related to neuronal hypoxia, neurogenesis, and neuronal metabolism and development. Particularly, miR-148b-3p upregulation promoted Schwann cell (SC) migration, whereas silencing of miR-148b-3p inhibited SC migration in vitro [43]. The molecular background of miR-148b-3p is in fact very interesting. It belongs to the miR-148/152 family [44], which includes miR-148a, miR-148b, and miR-152 and is considered a placenta-associated miRNA, which means it is expressed ubiquitously [43], not only in the placenta but also in other tissues. However, as indicated, the most interesting issue concerning its role in fetal medicine is its ability to promote the growth of SCs, responsible for myelin formation. miR-148-3p plays a role in the regeneration of peripheral nerves by regulating SC migration via targeting cullin-associated NEDD8-dissociated protein 1 (Cand1). Overexpression of miR-148-3p enhanced the migratory ability of SCs, while inhibition attenuated SC migration in vitro [43]. These effects occur in unison with other miRNAs such as miR-132, miR-210, miRNA sc-3, miR-221, and miR-222, which also increase the migratory ability of SCs, and miRNA sc-8, miR-9, miR-98, miR-1, and miR-182, which diminish this ability [45].
A parallel may therefore be drawn between peripheral nerve repair and axonal development (arborization) in the central nervous system. A good example of this is miR-132, which apart from promoting peripheral nerve repair mediated by SCs, as indicated, has been found to protect the central nervous system: miR-132 controls dendritic plasticity [46] and is required for normal dendrite maturation in newborn neurons [47]. Therefore, miR-132 functions as a key activity-dependent regulator of cognition, whose expression must be maintained within a limited range to ensure normal learning and memory formation [48]. In fact, miR-132 has been considered as a master regulator of neuronal health [49], and its supplementation is being evaluated for the treatment of diseases such as tau-associated neurodegenerative disorders [50]. Therefore, in an analogous way, miR-148b-3p might also play a role in the protection of the central nervous system. In theory, as brain tissue depends on myelination, miR-148b-3p might contribute to the protection of brain tissue under different circumstances, such as in chronic hypoxia [51].

Role in Energy Production
The possible role of miR-148b-3p in the protection of SCs might have a relationship to a number of biochemical functions. Regarding carbohydrates, miR-148b inhibits hypoxia-induced elevation of lactate production and hypoxia-induced increase in glucose consumption, thereby reducing cellular growth [52]. Regarding amino acids and proteins, miR-148b-3p and miR-25-3p behave as key regulators of biosynthesis of valine, leucine, and isoleucine and also regulate protein processing at the endoplasmic reticulum, both pathways of special relevance to fetal growth during the last trimester of pregnancy [35,36] and during periods of nutritional deprivation [53]. Finally, regarding fatty acid metabolism, both miR-25-3p and miR-148b-3p control biosynthesis of fatty acids and sphingolipids [31,34,54], essential molecules for stem cell differentiation morphogenesis and embryo development [54] that are also related to preeclampsia and FGR [55,56].

Clinical Implications
A practical result of the differential expression of miR-25-3p and miR-148b-3p in fetal blood would be the possibility to detect them also in maternal blood in order to develop clinical diagnostic tests. Hypoxia-related mi-RNAs produced in the placenta have been detected in maternal blood [57]. In this regard, if miRNAs are able to cross the placental barrier and circulate between the mother and the fetus [39], miR-148b-3p and miR-25-3p might also be detected in maternal serum and become markers of APO in an isolated or combined way, consequently improving the accuracy of late-onset FGR diagnosis.

Strengths and Limitations
The strengths of this study are, first, its novelty, as we have been the first investigators to perform full sequencing of all miRNAs in fetal blood, and, second, the finding of a miRNA profile directly related to neuronal development. Some shortcomings, however, might be the absence of validation in a different population, the paucity of follow-up data, and the absence of data related to neurocognitive evolution in childhood.

Conclusions
FGR fetuses express a different miRNA profile, which includes overexpression of miR-25-3p and, especially, miR-148b-3p, miRNAs related to cellular metabolism and neuronal plasticity. Future work is needed to assess the levels of miR-148b-3p and miR-25-3p in maternal serum in order to evaluate if they could improve the understanding and management of late-onset FGR, helping in the prediction of neurocognitive disability.