Maternal Age at Menarche Genes Determines Fetal Growth Restriction Risk

We aimed to explore the potential link of maternal age at menarche (mAAM) gene polymorphisms with risk of the fetal growth restriction (FGR). This case (FGR)–control (FGR free) study included 904 women (273 FGR and 631 control) in the third trimester of gestation examined/treated in the Departments of Obstetrics. For single nucleotide polymorphism (SNP) multiplex genotyping, 50 candidate loci of mAAM were chosen. The relationship of mAAM SNPs and FGR was appreciated by regression procedures (logistic/model-based multifactor dimensionality reduction [MB-MDR]) with subsequent in silico assessment of the assumed functionality pithy of FGR-related loci. Three mAAM-appertain loci were FGR-linked to genes such as KISS1 (rs7538038) (effect allele G-odds ratio (OR)allelic = 0.63/pperm = 0.0003; ORadditive = 0.61/pperm = 0.001; ORdominant = 0.56/pperm = 0.001), NKX2-1 (rs999460) (effect allele A-ORallelic = 1.37/pperm = 0.003; ORadditive = 1.45/pperm = 0.002; ORrecessive = 2.41/pperm = 0.0002), GPRC5B (rs12444979) (effect allele T-ORallelic = 1.67/pperm = 0.0003; ORdominant = 1.59/pperm = 0.011; ORadditive = 1.56/pperm = 0.009). The haplotype ACA FSHB gene (rs555621*rs11031010*rs1782507) was FRG-correlated (OR = 0.71/pperm = 0.05). Ten FGR-implicated interworking models were founded for 13 SNPs (pperm ≤ 0.001). The rs999460 NKX2-1 and rs12444979 GPRC5B interplays significantly influenced the FGR risk (these SNPs were present in 50% of models). FGR-related mAAM-appertain 15 polymorphic variants and 350 linked SNPs were functionally momentous in relation to 39 genes participating in the regulation of hormone levels, the ovulation cycle process, male gonad development and vitamin D metabolism. Thus, this study showed, for the first time, that the mAAM-appertain genes determine FGR risk.


Introduction
Fetal growth restriction (FGR) is defined as a pathological inhibition of fetal intrauterine growth and the inability of the fetus to reach its growth potential, in which fetal size is below the 10th percentile for a given gestational age [1,2].FGR is a common complication of pregnancy, the incidence of which reaches up to 10% worldwide [1,3].FGR is the leading cause of stillbirth (more than 50% of stillborn infants had FGR or were small for gestational age) and neonatal mortality/morbidity [4][5][6][7], and the long-term effects of FGR in adulthood are associated with an increased risk of developing cardiovascular/metabolic disorders (dyslipidemia; insulin resistance; arterial hypertension; obesity; type II diabetes; fatty liver, etc.) [8][9][10].
The FGR pathogenesis is complex and can be associated with various factors (maternal; placental; fetal), leading to a limited supply of nutrients and oxygen to the fetus [3,[11][12][13][14][15][16][17].Maternal risk factors such as advanced age, underweight, hypertension, diabetes, genetics, etc., are essential in the formation of FGR [2,18].The above violations of the maternal organism can facilitate the evolution of various placental disorders (vascular malperfusion disturbances in the maternal-fetal system (fibrin deposition/infarction/chronic abruption), chronic inflammation of placenta, etc.), leading to change in placental nutrient transport (glucose; amino acids; fatty acids; oxygen intake) which in turn can cause the formation of FGR [2,18,19].

FGR and FGR Free Group Characteristics
All basic medical/anamnestic/biological characteristics of FGR and FGR free groups from matched/unmatched data are shown in Table 1.According to the information presented in Table 1, such parameters as age (p = 0.017), pre-pregnancy maternal BMI (mBMI) parameter distribution (p = 0.0001), number of gravidity (p = 0.004) and induced abortions (p = 0.0001) in the anamnesis, the presence in anamnesis of arterial hypertension (p = 0.0001), FGR (p = 0.00001) and preeclampsia (p = 0.001) between the two cohorts compared were statistically different, and therefore these characteristics were included in the logistic regression models as confounders.

FGR-Significant Locus/Gene Probable Functions
In this section of the work, a detailed analysis of the functional relevance of 15 FGR-causal loci and 430 proxy SNPs was carried out, aimed at assessing the possible connection of the considered loci (n = 445) with missense mutations, epigenetic changes, gene expression and splicing regulatory effects and their involvement in FGR-related pathways.

Missense Mutations and FGR-Linked SNPs
Three loci determining the missense mutations were discovered by us among 430 proxy SNPs such as rs4889 KISS1 (replacement of proline with arginine in the 81 position of the KISS1 protein; "deleterious" SIFT grade), rs11676272 ADCY3 (replacement serine with proline in the 107 position of the ADCY3 protein; "benign" SIFT grade), rs61742688

FGR-Significant Locus/Gene Probable Functions
In this section of the work, a detailed analysis of the functional relevance of 15 FGRcausal loci and 430 proxy SNPs was carried out, aimed at assessing the possible connection of the considered loci (n = 445) with missense mutations, epigenetic changes, gene expression and splicing regulatory effects and their involvement in FGR-related pathways.

Missense Mutations and FGR-Linked SNPs
Three loci determining the missense mutations were discovered by us among 430 proxy SNPs such as rs4889 KISS1 (replacement of proline with arginine in the 81 position of the KISS1 protein; "deleterious" SIFT grade), rs11676272 ADCY3 (replacement serine with proline in the 107 position of the ADCY3 protein; "benign" SIFT grade), rs61742688 GPRC5B (replacement of asparagine with lysine in the 268 position of the GPRC5B protein; "benign" SIFT grade).

Syntropic Effects of mAAM-Involved Genes in FGR, mAAM, Pre-Pregnancy mBMI and Offspring BW
We conducted a comparative analysis of the data obtained in this paper with the results of our previous studies devoted to the study of associations of the same list of mAAM-related loci in the same population (women of the Russian ethnic group from Central Russia) with maternal AAM and BMI (mBMI) [36], BW [40], and uterus benign proliferative diseases (uterine leiomyoma (UL) [33]; endometrial hyperplasia (EH) [31]; endometriosis [32]) in order to identify common genetic factors underlying these reproductively significant phenotypes.As a result of this comparative analysis (Table S10), FGR-related polymorphisms associated with mAAM, mBMI and BW in this population were established (Table S10, [31][32][33]36,40]): out of 15 FGR-associated loci, 11 variants were associated with mBMI/BW (73.33%), including 5 SNPs with BW (33.33%), 2 SNPs-mBMI (13.33%) and 4 SNPs with both BW and mBMI (26.67%);only 2 FGR-associated loci were mAAM-associated (13.33%).The result outline of this study is presented in Figure 4.

Discussion
This is the first study to find the association of mAAM-involved gene polymorphisms and FGR susceptibility risk.Three SNPs independently (rs999460 NKX2-1; rs7538038 KISS1; rs12444979 GPRC5B), one haplotype FSHB gene (rs555621*rs11031010*rs1782507) and ten SNPs-interworking models of 13 loci were FGR-implicated.FGR-correlated mAAM-related 15 polymorphisms with 430 proxy variants were functionally meaningful to 39 genes participating in the regulation of hormone levels, the ovulation cycle process, male gonad development and vitamin D metabolism.
The present study revealed associations of rs12444979 GPRC5B with FGR both independently (the T allele of this SNPs was risky for FGR, OR = 1.56-1.67)and as part of SNP interactions models (50% of FGR risk models included this SNPs).Also, in our previously performed genetic study in this population (Russian women of Central Russia), the relationship of maternal rs12444979 GPRC5B (as part of interloci "dialogue" models) with offspring birth weight (BW) [40] was shown.rs12444979 GPRC5B was AAMbounded [41,42] and BMI-involved [43][44][45][46], wherein the T allele of this SNP correlated with a later AAM [41], and an alternative genetic variant for it, the C allele, was associated with increased BMI [43,44,46].Hence, the maternal AAM-boosting/BMI-lowering allele rs12444979 GPRC5B was an FGR risk factor.It is believed that low pre-pregnancy weight and low weight gain during pregnancy in a woman are linked to an increased FGR risk [11,15,17].
Our in silico materials pointed to significant SNP-eQTL and SNP-sQTL correlations: the FGR risk allele T rs12444979 GPRC5B was associated with higher KNOP1 gene mRNA production in multiple FGR-impact organs such as adipose (subcutaneous; visceral), skeletal muscle, ovary, thyroid, blood, adrenal gland, brain (hypothalamus, basal ganglia), lower GPRC5B transcription in subcutaneous adipose and low KNOP1 gene splicing level (together with 51 proxy SNPs) in adipose (subcutaneous), ovary, breast.The GPRC5B gene (it is placed on 12p13.1)encodes a receptor protein coupled with G proteins [47], playing an essential role in many different FGR-significant factors (insulin resistance, inflammation, cell growth/differentiation/apoptosis, etc.) due to intracellular signaling via pathways such as mitogen-activated protein kinase (MAPK)-c-Jun NH2-terminal kinase (JNK), transforming growth factor beta (TGF-β), interferon gamma (IFNγ), cyclic adenosine monophosphate (cAMP), nuclear factor κB (NF-κB), signal transducer and activator of transcription (STAT3), focal adhesion kinase (FAK)/Src family kinases (SRC), and signaling cascades [48][49][50].There are experimental data (obtained using transfected cell line models) on the key role of GPRC5B in metabolic stress processes (due to the modulation of interaction with phosphorylated sphingomyelin synthase 2) underlying lipid-generated insulin resistance [49].The fundamental value of GPRC5B in the development of a chronic inflammatory process and the formation of insulin resistance in adipose as a result has been proven on a GPRC5B-deficient mice model [51], which indicated the critical role of this gene in the regulation of insulin-susceptible organ (muscles, adipose, central nervous system) metabolism.These processes are essential in the pathophysiology of FGR [7].The KNOP1 gene (for which SNP rs12444979 GPRC5B was sQTL-significant) encoded a nuclear protein such as lysine-rich nucleolar protein 1 interacting with a zinc finger 106 protein [51].This protein has been involved in the regulation of particular developmental genes (e.g., TSG118, TSPYL) by chromatin rebuilding [52].
As a result of this study, correlation between rs7538038 KISS1 and FGR was shown (allele G, OR = 0.56-0.61).This SNP was functionally active (due to the KISS1 gene enhancers activity modulating) in FGR-correlated organs such as placenta, maternal brain (hippocampus, substantia nigra, cortex, etc.) and skeletal muscle, a multitude of fetal organs (brain male/female, muscle, kidney, lung, etc.) (our in silico materials).In previous studies, rs7538038 KISS1 was linked with AAM (the G allele was a genetic factor of early AAM) [53], central precocious puberty in girls (allele G was risky for central precocious puberty) [54], BW and endometrial hyperplasia (as part of inter-genic interworkings, [31,40], respectively).So, we can generalize that the AAM-reducing allele G rs7538038 KISS1, also associated with central precocious puberty of girls, is an FGR-protective factor.
The KISS1 gene encodes the kisspeptin protein (KP), which is cleaved into shorter, biologically active molecules-kisspeptins (KP-54, KP-14, KP-13; KP-10), which have a biological effect by activating the G-protein-coupled receptor 54 (GPR54), also known as the KISS receptor-1 [55].The KP/GPR54 system plays an important physiological role in neuroendocrine regulation of reproduction by influencing the hypothalamic-pituitarygonadal axis.In addition, it affects fertility, implantation processes, and stages of the menstrual cycle [56,57].In placental tissues, KISS1 mRNAs and KPs are found in the syncytiotrophoblast and, to a lesser extent, in the cytotrophoblast, whereas KISS1R is expressed in the syncytiotrophoblast and villous and invasive extracellular trophoblast [58].There are data on the implication of KPs in placentation regulation.In particular, KP-10 produced by trophoblast cells in the first trimester inhibits cell migration [58].It is assumed that a decrease in the expression of KISS1 and KISS1R may be correlated with a violation of placentation and FGR development: a number of studies have recorded a decrease in the level of KPs in pregnant women with FGR compared with women with physiological pregnancy [59][60][61].In addition, pregnant women with FGR are characterized by a weaker increase in the level of KPs throughout pregnancy and a lower level of KPs at the end of the first and third trimester of gestation [59].
The FGR-causal locus of rs999460 NKX2-1 increases this pregnancy complication risk (allele A, OR = 1.37-2.41)and determines susceptibility to FGR in various intergenic interactions.This SNP has a high functional potential: it was involved in the enhancer or/and promoter activity regulating disorder-linked cultured cells such as the derived CD184+_endoderm, the CD56+_ectoderm, the H1_BMP4_mesendoderm, etc., determining the nucleic acids binding with such 5TrFs as STAT, AIRE, Foxa, Arid5a, Pax-45 and controlling SFTA3 gene splicing in the thyroid (FGR-risk allele A was linked with high SFTA3 sQTL).Previously, associations of rs999460 NKX2-1 with AAM [53], female BMI [36], endometrial hyperplasia [31], and offspring BW [40] were presented.Interestingly, the AAM-boosting A allele rs999460 NKX2-1 [53] determined a high risk of FGR (OR = 1.37-2.41,our data) and was associated with a low BW [40].
NKX2-1 encodes a protein-factor transcription regulation (called TTF1 [thyroid transcription factor 1]) that communicates and "enables" (activates) promoters of the several thyroid-related hormone/protein genes such as thyroperoxidase, thyroglobulin, and the thyroid-stimulating hormone receptor [https://www.genecards.org/(accessed on 21 September 2023)].Besides this, the above hormones/proteins suppress TrF NR1D1 production and inhibit, due to this, the activity of genes important for processes of gluconeoand adipo-genesis, bile acid/lipid metabolism, and inflammatory cell responses [62].According to literature materials, transcription factor TTF1 (protein product of the NKX2-1 gene) is intensively synthesized in the process intrauterine development in several embryonic organs such as brain (hypothalamus, diencephalon, ventral forebrain, etc.), lungs, thyroid, etc. [63].Thyroid-implicated TTF1 production is vital for both the early stages of thyroid formation and embryonic development/growth overall [63].TTF1 with other several thyroid-associated TrFs (FOXE1/HHEX/PAX8) is collectively expressed in the process of thyroid formation in progenitor/mature follicular cells of this gland, providing thyroid formation/growth/differentiation/function/homeostasis at a necessary level [63].In TTF1 absence cases, thyroid progenitor cells may undergo apoptosis leading to their disappearance in the early stages of embryonic growth, resulting in a significant decrease in the number/mass of thyroid follicular cells; its formation is disrupted, and degradation occurs [64].The structural-functional disturbances of the thyroid may underlie various pregnancy complications such as premature childbirth, birth of preterm newborns and infants with low BW [63], which correlate with FGR outright [7].Epidemiological data indicate the presence of hypothyroxinemia at the 30th week of the gestation period in more than half of all infants with low BW [64].
Importantly, there is a unified result obtained for all three genes (KISS1, NKX2-1, GPRC5B) strongly associated with FGR: allelic variants of these genes are associated with an increased risk of FGR (the A allele, rs999460 NKX2-1 [OR = 1.37-2.41],and the T allele, rs12444979 GPRC5B [1.56-1.67]),which in previous studies showed a link with late menarche [41,53] and low BW (rs999460 NKX2-1 [40]) and adult BMI (rs12444979 GPRC5B [43,44]); and vice versa, the G allelic variant of the KISS1 gene (rs7538038) is correlated with a low risk of FGR (OR = 0.56-0.63),according to earlier studies, associated with the early menarche [53].It should be noted that the association of low maternal BMI with an increased risk of FGR is now believed to be proven and is not in doubt [2,11,15].There is also no doubt about the connection between genetic determinants of late/early AAM and low/high BMI, which has been repeatedly proven in previous research [34,35,[37][38][39].In accordance with the above scientific facts, the relationship we establish between the genetic factors of late/early mAAM and low/high FGR risk corresponds to the generally accepted ideas in this field at the present time: later menarche->low BMI->high FGR risk.
In this work, we convincingly demonstrate a significant similarity in genetic determinants (mAAM-related loci) FGR and BW/mBMI (73.33%) on the one hand and FGR and uterus benign proliferative diseases (66.67%) on the other hand.In our earlier study, the proportion of "common" polymorphisms for BW and mBMI (36.36%) and proliferative uterine diseases (UL (40.90%),EH (45.45%), endometriosis (36.36%) [40]) in general corresponded to the data of this study.Nevertheless, if the common genetic determinants for BW/mAAM/mBMI were four loci (rs1073768 GHRH; rs4374421 LHCGR; rs4633 COMT; rs4946651 LIN28B [62]), then in this study, no such "common" genetic factors were identified for FGR/BW/mAAM/mBMI, but "common" genetic variants for FGR/BW/mBMI were completely different from the above list of SNPs (4 SNPs-rs314280 LIN28B; rs3020394 ESR1; rs555621 FSHB; rs999460 NKX2-1).There are also significant differences in the list of "common" genetic determinants for FGR and three benign uterine diseases (3 SNPs FSHB-rs11031010; rs555621; rs1782507; materials of this study) and for BW and the same three uterine diseases (3 SNPs-rs12324955 FTO; rs4374421 LHCGR; rs1782507 FSHB; materials of the previous study [40]).These data show that despite the presence of "common" heredity (due to mAAM-related factors) between FGR and BW/mBMI/UL/EH/endometriosis and BW and mBMI/UL/EH/endometriosis, the specific genetic determinants (polymorphisms) underlying this differ significantly, despite the significant similarity of FGR and BW in the mAAM-linked polymorphisms that define them (60.00%).
The significant similarity in the genetic "architecture" of FGR and BW shown in our study (a high percentage, more than 50%, of common genes such as LIN28B; SFTA3; GC; ADCY3; LIN28B-AS1; LHCGR; ARL14EP; IQCK; GPRC5B; CENPO; ESR1; DNAJC27-AS1; EFR3B; FSHB; KNOP1; GPR139; HACE1; RBJ; KISS1; NKX2-1; REN; POMC; UGT2B4) is consistent with the biomedical logic (as a rule, a fetus/newborn with a low body weight is the "basis" of the FGR group) and the literature data on this topic [7,29].For example, the LIN28B gene (according to our data, this gene is essential for both FGR and BW) controls the formation of powerful specific regulators of the cell cycle (the let-7 family of miRNA) [65] and involved in the pubertal growth/development timing in girls/boys [66].There is persuasive experimental evidence that LIN28B-let7 regulating the insulin/phosphoinositide 3-kinases (PI3Ks)/mammalian target of the rapamycin (mTOR) pathway (by modulating the insulin receptor (INSR), insulin-like growth factor1 receptor (IGF1R), insulin receptor substrate 2 (IRS2) effects) is a midland controller of glucose metabolism (changing insulin resistance and glucose tolerance) in mammals [67].Several let-7 targets such as Hmga2, Myc, Igf2bp1, Kras are well-known controllers of glucose/insulin metabolism and mammalian size of body [68].The polymorphisms of this gene, according to the results of multiple investigations, have been associated with adult height/weight/BMI [69][70][71][72].Thus, our materials and literature data indicate that LIN28B may be one of the potentially relevant causal genes for both BW and FGR.

Study Design/Subjects
The outline of study design is shown in Figure 6.This case (FGR)-control (FGR-free) study included 904 women (273 FGR and 631 control) in the third trimester of gestation examined/treated in the Departments of Obstetrics (Belgorod Regional Clinical Hospital, Russia) during 2008-2017.All participants provided informed consent (form signed in person) prior to the start of this study.The medical ethics commission of both the Belgorod State University and the Belgorod Regional Clinical Hospital approved protocol/design of this study.
When forming the sample (FGR/FGR free), we used (a) a 24-41-week single pregnancy ending in a live birth, (b) Russian nationality (self-reported), (c) birthplace inside of Russia (Central region) [73,74] as inclusion parameters (criteria).Such parameters as age of <16 years, multifetal pregnancy, fetal/newborn congenital defects, maternal uterine congenital disturbance, and delivery at <24 weeks were used as exclusion criteria.
The participants of present research have previously been involved in other genetic studies of disorders/outcomes of pregnancy (newborn weight, preeclampsia, FGR) (detailed data on the results of these studies are contained in previous publications [40,75,76,[78][79][80][81][82]).Table 1 presents the biological/medical characteristics of the FGR/FGR free cohorts formed.
To collect 4-5 mL of peripheral (venous) blood, vacuum tubes (containing ethylene diamine tetra acetic acid (EDTA)) were used, from which genomic DNA was subsequently extracted (the "classical" method of isolation based on stepwise phenol-chloroform-ethanol procedures was used [129]).The resulting DNA was stored in kelvinator (temperature −80 • C).For multiplex genotyping on the Sequenom device (experimental genotyping procedures were carried out at the "Medical Genomics" Core Facility of Tomsk National Research Medical Center of the Russian Academy of Sciences (Tomsk, Russia), "working" samples with a DNA level (concentration) of 5-10 ng in one microliter were prepared (the Nanodrop-2000 measuring device (spectrophotometer) was used).To estimate the experimental data obtained, such indicators of SNP genotyping quality control were used as the call rate of at least 90% and the duplicate (blank) check success rate of at least 99% (90%) [33,130].One SNP, rs11724758 FABP2, did not meet the above quality requirements (call rate = 86.13%)and was excluded from further statistical genetic analysis.Overall, 49 SNPs corresponded to all of the abovementioned quality requirements.To collect 4-5 mL of peripheral (venous) blood, vacuum tubes (containing ethylene diamine tetra acetic acid (EDTA)) were used, from which genomic DNA was subsequently extracted (the "classical" method of isolation based on stepwise phenolchloroform-ethanol procedures was used [129]).The resulting DNA was stored in kelvinator (temperature −80 °C).For multiplex genotyping on the Sequenom device (experimental genotyping procedures were carried out at the "Medical Genomics" Core Facility of Tomsk National Research Medical Center of the Russian Academy of Sciences (Tomsk, Russia), "working" samples with a DNA level (concentration) of 5-10 ng in one microliter were prepared (the Nanodrop-2000 measuring device (spectrophotometer) was used).To estimate the experimental data obtained, such indicators of SNP genotyping quality control were used as the call rate of at least 90% and the duplicate (blank) check success rate of at least 99% (90%) [33,130].One SNP, rs11724758 FABP2, did not meet the above quality requirements (call rate = 86.13%)and was excluded from further statistical genetic analysis.Overall, 49 SNPs corresponded to all of the abovementioned quality requirements.

Statistical Genetic Analysis
Analysis of the HWE for every polymorphism in FGR/FGR free cohorts was performed [131,132].Association between FGR risk and mAAM-connected SNPs was detected by logistic regression with the help of such software tools as gPlink (version 1.07)

Statistical Genetic Analysis
Analysis of the HWE for every polymorphism in FGR/FGR free cohorts was performed [131,132].Association between FGR risk and mAAM-connected SNPs was detected by logistic regression with the help of such software tools as gPlink (version 1.07) [133] (for individual loci [four genetic models such as allelic, recessive, additive, dominant [134] were tested] and SNP haplotypes), MB-MDR (version 2.6) [135,136] and multifactor dimensionality reduction (MDR) [137,138] (for SNP interworking) taking into account multi-test calibration (permutation was performed) [139,140] and covariates (such as age, BMI before the current pregnancy, number of gravidity and induced abortions in the anamnesis, the presence in anamnesis of arterial hypertension, FGR and preeclampsia according to the information granted in Table 1).The following parameters, p perm , were declared as statistically meaningful: for individual SNPs, p perm ≤ 0.0125 (multi-test calibration based on Bonferroni correction [0.05/4 according to the quantity of examined genetic models] was performed); for SNPs haplotypes, p perm ≤ 0.050; for models of SNPs interactions, p perm < 0.001.It seems important that when choosing FGR-related SNP interworking models for permutation testing in order to obtain more reliable results, we used additional Bonferroni corrections (the potential quantity of 49-locus feasible recombination taken into account).As a result, significance level parameters p for models of different multi-locus levels were derived (used as "threshold indicators" for choosing FGR-related SNP interworking models for permutation testing) such as 2 SNP interworking -<0.05/1176 = 4 × 10 −5 ; 3 SNP interworking -<0.05/18,424 = 3 × 10 −6 ; 4 SNP interworking -<0.05/211,876 = 2 × 10 −7 [40].The subjects amounted to n = 904 (case = 273/control = 631), with a scheduled study power of ≥80% allowing identification of differences at the level of OR additive 1.33-1.56,OR dominant 1.59-1.63,OR recessive 1.61-4.71.Power indicators for FGR-linked loci were computed by the Quanto tool [141].

Conclusions
The present study proves the link between mAAM-involved gene polymorphisms with FGR mediated by functional effects of FGR-associated SNPs.The data obtained expand the understanding of the medico-biological significance of maternal age at menarche genes in the formation of pregnancy complications.

Figure 1 .
Figure 1.The entropy graph of SNP × SNP interactions with fetal growth restriction.The figure outlines the SNP×SNP interactions within the 2-, 3-, and 4-locus models obtained by the MB-MDR method.The polymorphisms are shown by the chromosome number and rs SNP ID.The percentage at the bottom of each SNP represents its entropy, and the percentage on each line represents the percentage of interaction between the 2 SNPs.The red and orange lines indicate a stronger and weaker synergism, respectively, brown-an independent effect of individual SNPs, green-weaker antagonism, blue-stronger antagonism.

Figure 1 .
Figure 1.The entropy graph of SNP × SNP interactions with fetal growth restriction.The figure outlines the SNP × SNP interactions within the 2-, 3-, and 4-locus models obtained by the MB-MDR method.The polymorphisms are shown by the chromosome number and rs SNP ID.The percentage at the bottom of each SNP represents its entropy, and the percentage on each line represents the percentage of interaction between the 2 SNPs.The red and orange lines indicate a stronger and weaker synergism, respectively, brown-an independent effect of individual SNPs, green-weaker antagonism, blue-stronger antagonism.

Figure 2 .
Figure 2. FGR-related protein-protein interaction networks inferred using the STRING resource.

Figure 2 .
Figure 2. FGR-related protein-protein interaction networks inferred using the STRING resource.

Table 1 .
Phenotypic characteristics of study participants.
Note: BMI, Body mass index; significant p-values showed in bold; *-regular and irregular (episodic) smoking at least 1 time or more per week; **-drinking low-alcohol drinks (wine, beer and others) or/and strong alcoholic beverages at least 1 time or more per week; p values < 0.05 are shown in bold.

Table 3 .
Associations of haplotypes with FGR.