Ancestry informative markers and selected single nucleotide polymorphisms in immunoregulatory genes on preterm labor and preterm premature rupture of membranes: a case control study

We conducted an ambispective case–control study of singleton pregnant women who delivered at Botucatu Medical School Hospital (Botucatu – São Paulo, Brazil) between 2003 and 2014. The aforementioned hospital is a tertiary center that provides assistance to 68 cities in the State of São Paulo (Southeast Brazil). The case group consisted of women with PTL with intact membranes or PPROM without other pregnancy complications. We collected buccal swabs from 157 women with PTL and from 114 of their babies, and from 80 women with PPROM and from 63 of their babies. Swab collection from PTL and PPROM patients was performed at their admission, while women were still pregnant. Since a significant number of patients did not live in the city and due to difficulties for collection of babies’ samples during their stay - short stays, newborns admitted into intensive care unit - it was not possible to collect samples from all the babies born to mothers included in the case group. We excluded 14 maternal and 23 babies’ samples that did not have sufficient material for analysis and 22 pairs of maternal and babies’ samples that were found to present systemic diseases (arthritis (1), hipertension (2)), fetal abnormalities (1) or gestational pathologies (preeclampsia (4), gestational diabetes (2), placenta previa (2), intrauterine growth restriction (3), intrauterine infection (2), oligoamnio (2), fetal distress (3)). The final case group consisted of 201 maternal samples (136 PTL and 65 PPROM) and 132 baby samples (88 PTL and 44 PPROM). For the control group we first collected 474 samples from mothers-babies pairs of healthy term deliveries, with no previous history of PTL or PPROM, and then matched the first 402 of these samples that met inclusion criteria and had sufficient material to the case group by newborn gender and maternal age – with a maximum difference of two years. Gestational age was calculated by the last menstrual period and confirmed by first trimester ultrasound. Discrepancies were corrected by the ultrasound result. Sociodemographic data, clinical information and personal and family histories were obtained through a standardized, closed-question questionnaire and by examination of medical records. The study was approved by the Human Research Ethics Committee from Botucatu Medical School (Protocol 3858–2011, FMB, Unesp). All patients enrolled provided written inform consent.

PTL with intact membranes was diagnosed as the presence of at least two regular uterine contractions every 10 min associated with cervical changes in patients with a gestational age between 20 and 37 incompleted weeks. Tocolysis was successfully achieved in 24.3 % of women from this group (33/136) who delivered after 37 weeks of pregnancy. PPROM was diagnosed by history and physical examination, which included documentation of nitrazine positive pooled vaginal fluid obtained by sterile speculum examination between 20 and 37 incompleted weeks of gestation. In this group only 3 women (4.6 %) had their pregnancies prolonged over 37 completed weeks.

Genomic DNA was extracted from buccal swabs in automated Qiacube equipment using QIAamp® DNA Mini Kit (Qiagen) and DNA quantification was performed by spectrometry using Epoch (Biotek). We evaluated 17 SNPs related with eleven different immune modulatory genes as described in Table 1. The SNPs were genotyped using Taqman® SNP Genotyping Assays and Taqman® Genotyping Master Mix (Applied Biosystems) following manufacturer recommendations. PCR reactions were run in 7500 Real-Time PCR System (Applied Biosystems). As there was no Taqman® assay available to genotype rs3918242 in the MMP9 gene, the identification of this SNP was performed by PCR-RFLP (Restriction Fragment Length Polymorphism) using the primers MMP9 F 5’- GCC TGG CAC ATA GTA GGC CC-3’ and MMP9 R 5’- CTT CCT AGC CAG CCG GCA TC-3’ for amplification, followed by incubation with the restriction enzyme SphI (Biolabs) [40]. For every SNP evaluated, three clinical samples initially identified by real-time PCR or PCR as wild-type homozygous, mutated homozygous and heterozygous were subjected to direct sequencing and used as positive controls to guarantee the reliability of results. Two negative controls (sterile water) were also included in each run. The reproducibility of results – obtained by repeating 10 % of the samples randomly chosen for all assays – was 99.98 %.

We designed pairs of primers for all SNPs to be evaluated (Table 1). After DNA amplification the samples were quantified in agarose gel by comparison with Low DNA Mass Ladder (Invitrogen) and purified using Illustra™ GFX™ Gel Band Purification Kit (GE Healthcare). The purified DNA sample was then amplified using reverse and forward primers separately with BigDye® Terminator v3.1 and, subsequently to the steps of DNA precipitation and resuspension, sequencing was performed in 3500 Hitachi equipment (Applied Biosystems). Sequences were analyzed using BioEdit.

For the identification of maternal Ancestry Informative Markers, a panel of 61 selected insertion/deletion (indel) variable sites were amplified in conditions as described by Resque et al. [44]. The indel markers that constitute this panel were selected based on the characteristic of exhibiting substantially different frequencies between population from different geographic regions. Then the samples were genotyped using ABI PRISM® 3130 Genetic Analyzer (Applied Biosystems) and the results analyzed using the GeneMapper v3.2 (Applied Biosystems). The ladder ABIGS LIZ-500 (Applied Biosystems) was used as a reference for the identification of each indel. A standard of known size was included in each run to ensure quality control of the analysis. As the admixture model assumes that each individual inherits part of their ancestral markers from ancestral populations, the results were plotted against the three parental populations from our database [45] that constitute the Brazilian population – Amerindian, Western European and Sub-Saharan African – to perform ancestry stratification. For twelve samples from the case group there was not enough material to perform this analysis. The software Structure v2.3.4 with 50,000 burnin length was used to estimate admixture.

Allele, genotype and haplotype frequencies were obtained by direct counting. Linkage Disequilibrium (LD) and expectations under the Hardy-Weinberg proportions were evaluated using Genepop 3.4. Haplotypes were inferred using the PHASE algorithm with final iteraction increased 10 times [46]. Allele and haplotype frequencies were compared by Fisher’s exact test and Odds Ratio and genotype frequencies by χ ² test and Odds Ratio. Maternal allelic data was adjusted by ancestry and smoking by logistic regression using stepwise backwards. Clinical and sociodemographic data were analyzed by Fisher’s exact test and Mann–Whitney. Additionally, in order to perform a more complete evaluation of the SNPs associated to adverse outcomes, their frequencies in different populations were obtained from the 1000 genomes database [47], the haplotypes were inferred using the software Arlequin 3.5 and the frequencies were compared among populations. The software used were GraphPad® Prism 5.0 and SAS 9.3. A p-value <0.05 was considered statistically significant.

Sociodemographic data are displayed in Table 2. Marital status, self-reported ethnicity, parity and years of education were similar between the groups. Family history of PTL and/or PPROM was less common in the control group when compared to PTL (p < 0.001) or PPROM (p < 0.001). Smoking was increased among women with PTL (p = 0.007) when compared to controls. Newborns of PTL and PPROM mothers had significant lower birth weight and apgar scores then those born at term (Table 3).

Median of European, Amerindian and African ancestries from women included in the study are shown in Table 4. European ancestry was increased among PTL when compared to controls (p = 0.002) while African ancestry was higher in controls when compared to PTL (p = 0.009).

Genotype frequencies were under Hardy-Weinberg expectations. We detected associations concerning allele or genotype frequencies for TLR2, IL10 and TNFA genes and the studied complications. Regarding IL1B, IL6, IL6R, MMP9, TNFR, TLR4, TIMP1 and TIMP2 genes, however, there were no differences between PTL or PPROM and controls.

In maternal samples no alleles or genotypes were associated to PTL when compared to controls. When we compared controls with PPROM, the alleles TLR2A (rs4696480) (p = 0.007) and TNFA-238G (p = 0.009) (Table 5) as well as the genotypes TLR2 AA (p = 0.004) and TNFA-238 GG (p = 0.012) (data not shown) were associated with this complication . Regarding the babies’ alleles, IL10-592C (p = 0.01) and IL10-819C (p = 0.026) were more frequent in PTL than in controls (Table 6). There was no difference in allele frequencies between PPROM and controls in babies' samples for any SNP evaluated.

The test of genotypic disequilibrium indicated the presence of LD among IL1B, IL6R, IL10, TLR4 and TNFA SNPs (p < 0.001 for all). Given that, haplotypes were inferred by probabilistic models as described in methods section. In maternal samples, the haplotype TNFA-308G-238A was protective against PTL (p < 0.001) when compared to controls (Table 7). No association was found between the studied phenotypes and haplotypes in PPROM group or in the babies’ samples. No haplotypes in IL1B, IL6R, IL10 and TLR4 genes were associated with PTL or PPROM.

We used logistic regression in maternal data to postulate different models to analyze the effect of polymorphisms, ancestry and smoking combined. European contribution and smoking increased the risk for PTL while African ancestry was protective against this outcome (Table 8). Regarding PPROM, presence of IL10-1082G and TLR2A (rs4696480) increased the risk for this complication while the allele TNFA-238A was protective (Table 9).

We also used logistic regression to evaluate whether the time interval between PTL or PPROM initiation and time to delivery (TD) was influenced by the variables SNPs, genetic ancestry and smoking. For this analysis we only considered women with PTL or PPROM and splitted them into two groups: short TD (≤24 h) and long TD (>24 h). Women positive for the allele IL1B-31T were at higher risk for short TD while those carrying TLR4-299G had a longer TD. Genetic ancestry and smoking did not influence this parameter (Table 10).

Lastly, we thought to compare women who delivered very preterm infants (≤34 weeks of gestation) with those with late preterm neonates (>34 weeks of gestation). For this comparison we only included women that delivered prematurely (births initiated either by PTL or by PPROM) and separated them into very preterm and late preterm subgroups. None of the variables - polymorphisms, ancestry or smoking - could be used to create a model to differentiate very preterm from late preterm subgroups.

European ancestry and smoking increased the odds of PTL while African ancestry was protective. The presence in babies of alleles IL10-592C and IL10-819C was also associated with PTL. Maternal presence of IL10-1082G and TLR2A (rs4696480) increased the risk for PPROM while TNFA-238A was protective. No fetal alleles were associated with PPROM, possibly because of the small size of this subgroup. Regarding haplotypes, TNFA-308G-238A was protective against PTL in maternal samples. Family history of PTL and/or PPROM was also associated with these outcomes, and time to delivery was influenced by the presence of IL1B-31 T and TLR4-299G.

One limitation of our study is that this is a single center study, which, despite the positive effect on homogeneous sampling, may include bias such as social background. The strength is the analysis of ancestry markers in admixed population to stratify ethnicity by unbiased methodology for the first time in the literature.

Populations are generally mixed, and the Brazilian population is one of the most heterogeneous in the world. As self-reported ethnicity is a poor predictor of genomic ancestry [48, 49], we evaluated AIMs to access the role of ancestry in the studied phenotypes. We observed higher contribution of European-originated markers in the PTL group, which at first seemed unexpected once the literature reports higher rates of this complication in African-descendant women [12, 13]. However, these studies are mainly originated from the USA or Europe, regions with different environments than Brazil. As frequencies of polymorphisms are unevenly distributed among populations, the higher European ancestry among our PTL patients may reflect higher frequencies of SNPs that increase the risk for this outcome in the Southeast Brazilian environment. It has been hypothesized that variations of exposure to microorganisms in distinct ancestral environments may have resulted in a selective pressure that maintained genetic variants that increase survival in response to infectious stimuli [12]. In a new environment, polymorphisms that were advantageous may become detrimental [12, 50]. Thus, one specific allele may induce different responses, i.e. confer resistance or susceptibility, in distinct environmental backgrounds. In this way, more studies to identify such polymorphisms in our population are needed. To date, we are the first to use such unbiased approach to evaluate the influence of ancestry in PTL and PPROM in mixed population.

We also described an increased risk for PTL in women who smoke. Exposure to tobacco during pregnancy is a well-documented risk factor for pregnancy complications [51, 52] and increases the risk for fetal morbidities [53]. Preterm infants are more likely to be born to mothers who smoke [54]. Indeed, the risk of preterm birth attributable to smoking has been estimated as more than 25 % [55]. Chang et al. [56] suggested cessation of smoking during pregnancy as a part of a strategy to reduce preterm birth rate in developed countries.

Regarding PPROM, the presence of IL10-1082G and TLR2A increased the risk for this outcome. Interleukin 10 (IL-10) is a potent regulator of inflammatory response and altered levels of this mediator are involved in the pathophysiology of PTL and PPROM. Nevertheless, there are contradictory findings regarding the influence of polymorphisms located in its promoter region in the IL-10 expression. Annells et al. associated the low producing haplotype IL10-1082A-819 T-592A to the inflammatory events of delivery before 29 weeks of gestation [57] and risk of chorioamnionitis [41] while other authors did not find association between these SNPs and adverse pregnancy outcomes [31, 43]. On the other hand, the high producing IL10-819C and IL10-1082G alleles have also been implicated in the etiology of complications with an inflammatory signature such as preeclampsia [58], and even delivery before 29 weeks of pregnancy [59], and there are reports that correlate the IL10-1082A-819 T-592A haplotype with a reduced risk for small-for-gestational age [60]. In the present we report the association between maternal IL10-1082G and PPROM and between presence of IL10-592C and IL10-819C in babies and PTL. The presence of these alleles may disrupt the balance between pro- and anti-inflammatory cytokines, increasing the risk for PPROM and PTL. It is also worth considering that some of the associations found between SNPs and diseases in genetic studies may be spurious as, as mentioned before, they may reflect differences in the distribution of SNPs in distinct populations and as such are risk markers rather than risk factors. For instance, the allele IL10-592C reported here to be more frequent among children born preterm is more common in European populations [47], this type of ancestry was also reported by us to increase the risk for this complication.

Toll-like receptors (TLR) are transmembrane proteins that recognize pathogen-associated molecular patterns and play an essential role in innate immune responses. TLR signaling positively regulates the expression of pro-inflammatory genes. Genetic variants in TLR pathways may alter the susceptibility to early PTL and to neonatal complications such as sepsis and necrotizing enterocolitis in preterm newborns [61, 62]. Our study is the first to report an association between a SNP at TLR2 and PPROM. Sutherland et al. [63] examined the same SNP in patients with sepsis and observed an association between the A allele and development of sepsis and Gram-positive cultures. Considering the importance of subclinical infection in the setting of PPROM it is possible that the presence of the A allele at position rs4696480 facilitates intraamniotic colonization by gram-positive bacteria activating the inflammatory pathways that culminate in PPROM.

The pro-inflammatory cytokine TNF-α also plays an important role in PTL and PPROM. TNFA-308A increases the production of TNF-α and has already been associated with PTL and PPROM [31, 32, 64]. Liang et al. [64] suggested that the presence of at least one TNFA-308G allele could be protective against PTL. Consistent with the literature, we report a protective role for the low-TNF-α-producing haplotype TNFA-308G-238A, located in the promoter region, against PTL and association between the low-producing TNFA-238A and decreased risk of PPROM.

Another interesting finding is the association between IL1B-31T and short TD and TLR4-299G and increased TD. The mutated IL1B-31T results in 2-fold increased mRNA production compared to the ancestral allele [65]. Thus, as this variant can lead to increased IL-1β levels, the association between the T allele and short delivery latency is perfectly plausible. Conversely, the TLR4-299G allele reduces the responsiveness to LPS in cell cultures [66]. Once the pro-inflammatory cascade triggered by TLR-4 binding is less efficiently induced in these patients they are likely to present longer delivery latency. After the PTL and PPROM pathways are triggered tocolytic treatments currently available are mostly inefficient. In a study of women with an unfavorable cervix who required labor induction, Doulaveris et al. [67] reported that patients with the GG genotype in the ATG16L1 gene - associated with a decreased capacity to induce autophagy - had a reduced time to delivery. The authors suggest this finding may result from increased production of IL-1β due to impaired autophagy. The identification of specific alleles that contribute to the determination of time to delivery can be of value for clinical practice. Finally, the association between family history of PTL/PPROM and their re-occurrence in the study group reinforces the role of genetic alterations in these outcomes.