The role of FAS, FAS-L, BAX, and BCL-2 gene polymorphisms in determining susceptibility to unexplained recurrent pregnancy loss

Abstract

Purpose

Idiopathic recurrent pregnancy loss (RPL) is a multifactorial reproductive disorder where an impaired control of apoptosis is likely involved. Triggering the cell death mechanism occurs in a spatiotemporal manner and is strongly related to a healthy pregnancy. Single nucleotide polymorphisms (SNPs) at the regulatory regions of genes are known to influence the expression patterns of apoptosis-related molecules.

Methods

A total of 296 unrelated female Brazilian patients were evaluated for clinical-demographic variables and genetic factors: 140 women who had experienced an unexplained RPL (with at least two consecutive abortions) and 156 healthy multiparous women. In all patients, six SNPs were evaluated in genes of apoptosis-related pathways: FAS (rs2234767, rs1800682), FAS-L (rs763110, rs5030772), BAX (rs4645878), and BCL-2 (rs2279115) by PCR followed by a restriction fragment length polymorphism (RFLP)-based analysis.

Results

The BAX-248GA genotype is independently associated with idiopathic RPL [adjusted OR = 0.30, 95% CI 0.13–0.70, P = 0.005] susceptibility. In the same multivariate model, the variables ethnicity, smoking, and alcohol consumption were statistically associated with RPL susceptibility (P < 0.05). No association with RPL susceptibility was reported for the remaining SNPs.

Conclusion

Our study is the first to evaluate the role of the main SNPs from both the extrinsic and intrinsic apoptosis pathways in RPL susceptibility. The association of BAX-248G/A with RPL susceptibility suggests that maternal predisposition for RPL has an essential contribution from genes involved in the delicate balance of endometrium cell turnover (cell death/proliferation). Therefore, apoptotic genes may represent promising targets for future studies on healthy pregnancies and the spectrum of pregnancy disorders.

Electronic supplementary material

The online version of this article (10.1007/s10815-019-01441-w) contains supplementary material, which is available to authorized users.

Introduction

Recurrent pregnancy loss (RPL) is a complex pregnancy disorder that is characterized by women who have experienced three or more miscarriages. Currently, some guidelines accept the threshold of having two or more consecutive pregnancy losses until 24 gestational weeks (GWs) [1]. RPL incidence may vary according to the diagnostic criteria. When defined by the presence of at least two consecutive miscarriages, RPL affects 2–5% of couples, and for three miscarriages, the incidence is ~ 1% [2, 3]. RPL can occur at two following subclinical levels: primary and secondary. Primary RPL refers to women who have never had a successful pregnancy; secondary RPL refers to women who have had a successful pregnancy at term with a live newborn, followed by having a series of miscarriages [4]. Although some risk factors are associated with the disorder, the etiopathology of RPL is still unknown [2]. After the identification of the potential causes that lead to RPL, approximately 50% of recurrent miscarriage cases remain idiopathic [5].

The maintenance of normal pregnancy requires a complex network of interactions and the involvement of distinct, but correlated, biological processes [6, 7]. In this context, the differential expression of genes in the endometrial tissue and in the chorionic villi in normal pregnancies and RPL has been observed [8–12]. The variable results of the comparative gene expression that is seen in the literature reflect the heterogeneity of studies since the cellular composition of tissue samples from RPL and normal pregnancies may differ. Nevertheless, such studies provide important clues for potential biomarkers or therapeutic targets to improve the clinical management of women with RPL. In this sense, the higher expression of apoptosis-related genes, such as the pro-apoptotic genes FAS, FAS-L, BAD, BAX, BID, and the CASPASE family (3, 6, 7, 8, 9, 10, and 12), was observed in the chorionic samples from women with RPL compared with samples from normal pregnancies [8]. In contrast, the expression of antiapoptotic genes, such as BCL-2 and BCL-XL, was similar among chorionic samples from RPL and normal pregnancies, suggesting that apoptosis-related genes are directly involved in RPL [8]. In addition, it has been shown that apoptosis levels are higher in chorionic samples from women with RPL than in samples from electively terminated first-trimester pregnancies [13].

Apoptosis is fundamental to human reproduction because this phenomenon is directly involved in endometrial cell turnover, spiral artery remodeling, the development of the embryo, and in maternal-placental immunologic tolerance [4, 14]. Once apoptosis signaling is triggered by either the extrinsic (death receptor-mediated apoptosis) or the intrinsic (mitochondrial) pathways, it usually results in cell death by the activation of the caspase cascade. The main molecules of the extrinsic pathway are the death receptor FAS and its ligand FAS-L [15]. FAS is constitutively expressed in most cell types. FAS-L has a restricted tissue distribution, for example, in the testes, the eyes, activated T lymphocytes, and natural killer cells [16]. FAS/FAS-L signaling contributes to the maintenance of the immunologically privileged status of some tissues, including the maternal-fetal interface. It has been reported that extravillous trophoblast cells (EVTs) express FAS-L [17], thus implying that FAS-L expression in both maternal and fetal cells represents an additional mechanism related to the immunotolerance phenomenon in pregnancy [18]. Genetic variants in the regulatory region of FAS and FAS-L genes have been shown to impact gene expression patterns and the susceptibility of many immune-related disorders and pregnancy disorders, including RPL [19–27]. FAS-670A/G and FAS-1377G/A are single nucleotide polymorphisms (SNPs) that are located in the promoter region of FAS. They are in the binding site of critical transcription factors: the signal transducer and activator of transcription-1 (STAT-1) and specificity protein-1 (SP-1) [28, 29]. Two SNPs in the FAS-L have also been identified: FAS-L IVS2nt-124A/G, located in intron 2, and FAS-L-844T/C in the promoter region. The latter modifies the affinity for the CCAAT/enhancer-binding protein beta (C/EBPbeta). In addition, the FAS-L-844C allele confers higher FAS-L expression in lupus erythematosus patients [30].

The intrinsic pathway is under the control of proteins belonging to the BCL-2 family, with BCL-2 and BAX being the significant players of this pathway. Interestingly, apoptotic stimuli, such as oxidative stress, DNA damage, or hypoxia, lead to the activation of the downstream signaling of the intrinsic pathway [31]. BCL-2 has antiapoptotic activity and prevents cell death either by sequestering inactive forms of death-driving caspases or by regulating the release of mitochondrial apoptosis factors, such as cytochrome c, which binds to apoptotic protease-activating factor 1 (APAF-1) to form the apoptosome in the cytoplasm [32]. In contrast, when BAX is upregulated, it counteracts the negative effects of BCL-2 and promotes the release and activation of caspases through the regulation of mitochondrial permeability [33].

The BCL-2 gene has two promoter regions, named promoter 1 (P1) and promoter 2 (P2). P1 has a stimulatory function, and P2 has an inhibitory function. If P2 activity is higher than that of P1, it decreases gene transcription, which is a negative regulatory element. The SNP located in P2-938C/A is reported to influence the binding of the transcription factor SP-1. Nuckel et al. [34] reported that in B cell chronic lymphocytic leukemia (CLL) patients, the P2-938AA genotype has a higher BCL-2 expression. Additionally, in transfected cells (specifically HEK293 and Karpas 422), the BCL-2 P2-938C allele results in the reduced transcriptional activity of the gene due to the increased activity of P2. However, the functional relevance of this SNP on endometrial tissue and pregnancy disorders remains to be investigated.

Currently, it is suggested that BAX-248G/A has no influence on cancer susceptibility [35, 36]. The effect of SNP BAX-248G/A on the expression levels of the protein in CLL patients is still controversial [37, 38]. However, initial reports observed that the BAX-248A allele is associated with low expression levels in untreated CLL patients [39]. In addition, this SNP has been suggested to influence drug (fludarabine) responses in CLL patients and provides insights for pharmacogenomic studies in other clinical settings [39, 40].

In pregnancy, previous studies have reported the high expression of apoptotic proteins in the chorionic villi from women with RPL [8, 41], thus suggesting that imbalances between the proapoptotic and antiapoptotic stimuli are likely involved in RPL pathogenesis. Currently, studies evaluating the role of genetic variants in FAS, FAS-L, BAX, and BCL-2 on RPL susceptibility [21, 22] are limited or even absent, and in most cases, a limited set of variants from a single apoptotic pathway was evaluated. Therefore, we evaluated whether SNPs in critical genes of the classical apoptosis pathway are associated with RPL risk in a Brazilian population.

Materials and methods

Selection and description of participants

We recruited patients at the Prenatal Diagnosis Clinic of the Medical Genetics Service at the Hospital de Clínicas de Porto Alegre (HCPA), Brazil. The current case-control study enrolled 140 idiopathic women with RPL: 101 (72.1%) primary RPL patients and 39 (27.9%) secondary RPL patients. The inclusion criteria comprised women reporting at least two pregnancy losses before 24 GWs with the same partner. A structured interview was performed to obtain the clinical and demographic data and the obstetric history. Before inclusion in the study, all patients underwent a standardized clinical evaluation, which included hysteroscopy, laparoscopy, ultrasound, and a full determination of hormonal status to identify the cause of the pregnancy losses. Karyotype examinations evaluated the chromosomal abnormalities (numerical and structural). Immunological risk factors were also investigated through the assessment of anticardiolipin, lupus anticoagulant, and antinuclear antibody levels; the patient medical history of autoimmune diseases was also reviewed. We considered the presence of any maternal clinical condition that could prevent a full-term pregnancy as exclusion criteria for this study. Alcohol consumption and cigarette smoking were defined as regular consumption/utilization (yes/no).

The control group included 156 women with two or more successful pregnancies (healthy multiparous women) and no history of pregnancy loss or infertility. These women were randomly selected to participate in the study during blood collection for routine laboratory analyses at the Hospital de Clínicas in Porto Alegre (HCPA). These women verbally answered a standardized questionnaire about demographics, consanguinity, family history, and the number of pregnancies. The results were reported and analyzed following the previous recommendation by STREGA [42].

Genomic DNA extraction

Biological samples consisted of blood or saliva. The genomic DNA from the blood samples was extracted from leukocytes according to the methods by Lahiri and Numberger [43] and was maintained at − 20 °C. The DNA from the saliva samples was obtained using the Oragene® DNA collection kit (DNA Genotek Inc., Canada) according to the manufacturer’s protocol.

Polymorphism genotyping

The primer sequences, polymerase chain reaction conditions, and the genotyping methods for the polymorphisms of FAS-L-844T/C (rs763110), FAS-670A/G (rs1800682), and FAS-1377G/A (rs2234767) have been previously described by Sun et al. [44]. The amplification reaction and the genotyping conditions for FAS-L IVS2nt-124A/G (rs5030772) were previously described by Zhang et al. [45]. The BAX-248G/A (rs4645878) and BCL-2-938C/A (rs2279115) variants were analyzed as described, respectively, in Moshynska et al. [38] and Zhang et al. [46]. Briefly, the PCR products were digested with the FokI, BsrDI, ScrFI, BstUI, AciI, and BccI restriction enzymes (New England Biolabs, Beverly, MA) to distinguish the FAS-L IVS2nt-124A/G, FAS-L-844T/C, FAS-670A/G, FAS-1377G/A, BAX-248G/A, and BCL-2-938C/A polymorphisms, respectively. All genotypes were determined by electrophoresis with an 8% polyacrylamide gel stained with ethidium bromide. Finally, we randomly selected 10% of the samples for an independent repetition of the genotyping procedure, and the results were 100% concordant with the initial data.

Statistical analyses

Chi-squared and Mann-Whitney tests were used to compare categorical and continuous variables, respectively. The median and the 25th–75th percentiles describe asymmetric distributions. The Hardy-Weinberg (HWE) equilibrium was calculated in the control samples using web-based software: http://www.oege.org/software/hwe-mr-calc.shtml [47]. The expectation-maximization algorithm determined the presence of linkage disequilibrium in the FAS and FAS-L genetic variants using Haploview software [48]. Analyses were stratified according to the RPL subclinical types and were compared with the control group. Binary logistic regression analyses were performed to assess the strength of the association between the genetic markers and the incidence of RPL by evaluating the odds ratio (OR) and the confidence interval (CI). Covariates were included in the multiple regression models based on the cutoff values for significance [association with the study factor (genotype) and with the final significance of P < 0.20]. The Bonferroni correction was applied for multiple tests, and the alpha was set at 0.0083 (α_Bonferroni = 0.05/6 SNPs tested). For all other nonindicated instances, the nominal P value < 0.05 was considered to be statistically significant. All statistical analyses were performed with standard software (SPSS for the Social Science, v.18.0 for Windows).

Results

In total, we evaluated 140 cases: 101 (72.1%) were diagnosed with primary RPL, and 39 (27.9%) were diagnosed with secondary RPL. The demographic and clinical characteristics of the individuals, as well as the pairwise comparisons between the RPL clinical subtypes and the controls, are described in Table 1. Smoking prevalence was not different among the studied groups (P > 0.05). In addition, the prevalence of smoking in controls (21.8%) was similar to the general estimates for pregnant women (17.5%) from the major cities of Brazil [49]. A high frequency of European ancestry (77.6%) was observed in the control group. This observation is in accordance with a sampling of the Brazilian population (82%), and European ancestry is the most frequent ethnic group in the southern region of Brazil [50]. Alcohol consumption was more frequent in women with primary and secondary RPL than in healthy multiparous women (P < 0.001 and P = 0.043, respectively).

Table 1.

Comparison of clinical and demographic characteristics among idiopathic RPL women and healthy multiparous women

Variables	Primary RPL (n = 101)	Secondary RPL (n = 39)	Healthy multiparous women (n = 156)	Primary RPL vs. secondary RPL	Primary RPL vs. healthy multiparous women	Secondary RPL vs. healthy multiparous women
				P value
Age at first pregnancy, median (25–75%)	23 (18.0–29.0)	20 (18.0–25.0)	22 (20.0–25.0)	0.113	0.224	0.212
BMI, median (25–75%)	25 (22.5–28.5)	26 (24.0–29.0)	24.0 (21.0–28.0)	0.259	0.226	0.057
European-derived, n (%)	59 (58.4)	29 (76.3)a	121 (77.6)b	0.051	< 0.001*	0.763
Alcohol consumption, n (%)	46 (45.5)	14 (35.9)	32 (20.5)	0.301	< 0.001*	0.043*
Smoking, n (%)	14 (13.9)	7 (17.9)	34 (21.8)	0.544	0.111	0.598
Alcohol and smoking habit, n (%)	11 (10.9)	4 (10.3)	9 (5.8)	1.000	0.134	0.297
Number of pregnancies, mean (±SD)	3.4 (1.8)	4.41 (1.39)	2.87 (1.15)	N/A	N/A	N/A
Miscarriages, mean (±SD)	3.4 (1.8)	2.97 (1.06)	N/A	N/A	N/A	N/A
2 miscarriages, n (%)	36 (35.6)	14 (35.9)	N/A	0.151c	N/A	N/A
3 miscarriages, n (%)	32 (31.7)	18 (46.2)	N/A		N/A	N/A
> 4 miscarriages, n (%)	33 (32.7)	7 (17.9)	N/A		N/A	N/A

RPL recurrent pregnancy loss, BMI body mass index, SD standard deviation, N/A not applicable

^aOne patient has missing data

^bTwo patients have missing data

^cP value of categorical miscarriages between primary vs. secondary RPL

*Statistically significant

The control group adheres to HWE expectations for all SNPs evaluated. Linkage disequilibrium (LD) was evaluated for FAS (-670A/G and -1377G/A; D′ = 1.0, r² = 0.111) and FAS-L (FAS-844T/C and IVS2nt-124A/G; D′ = 0.89, r² = 0.10). No differences in the haplotype frequencies between RPL and control patients were observed (Supplementary Table 1).

Multiple logistic regression models were performed to assess the effect of individual genotypes on primary RPL risk. A regression model adjusting for the effects of covariates and BAX-248G/A genotypes is presented in Table 2. The BAX-248GA genotype (OR = 0.30, 95% CI 0.13–0.70, P = 0.005) was independently associated with protection for primary RPL. The covariates ethnicity (OR = 0.42, 95% CI 0.23–0.75, P = 0.004), smoking (OR = 0.45, 95% CI 0.21–0.99, P = 0.049), and alcohol consumption (OR = 3.87, 95% CI 2.08–7.16, P < 0.001) were also statistically significant in the model. Additionally, we highlight the fact that the BAX-248AA homozygous genotype was exclusively observed in women with primary RPL.

Table 2.

Multivariate logistic regression for the effect of BAX-248G/A genotypes in idiopathic primary RPL corrected for covariates

SNP	Primary RPL, n (%)a	Healthy multiparous women, n (%)b	OR	95% CI	P
BAX-248 G/Ac
GG	87 (87.0)	123 (79.4)	Ref.
GA	10 (10.0)	32 (20.6)	0.30	0.13–0.70	0.005*
AA	3 (3.0)	–
Ethnicity	59 (58.4)	121 (78.6)	2.39	1.32–4.34	0.004*
Smoking	14 (13.9)	34 (21.8)	0.46	0.21–0.99	0.049*
Alcohol consumption	46 (45.5)	32 (20.5)	3.87	2.08–7.16	< 0.001*

Genotypes were not available for one patient in RPL and one in control. Reference categories were European-derived (ethnicity), “no” (smoking) and “no” (alcohol consumption)

OR odds ratio, CI confidence interval, Ref. reference category, P P value

^aPrimary RPL, n = 101

^bHealthy multiparous women, n = 156

^cOR for BAX-248AA genotype is absent due to a low count AA genotype in controls

*Statistically significant values

When the FAS, FAS-L, and BCL-2 polymorphisms and primary RPL were evaluated individually, there was no correlation even when the effects of genotypes were adjusted for covariates or in dominant and additive genetic models (Supplementary Table 2). In addition, no significant differences in polymorphisms between the secondary RPL and control cases were observed (Supplementary Table 3). Additionally, no significant difference in the pooled analysis of RPL clinical subtypes compared with the control subtypes was observed (data not shown).

Discussion

Few studies have evaluated the effects of FAS and FAS-L SNPs on the risk of RPL [21, 22, 27]. In an Indian population, Nair et al. (2012) assessed the impact of FAS-670A/G and FAS-1377G/A SNPs on RPL risk. In this study, they observed an association of FAS-1377A allele carriers with RPL risk. Nevertheless, the combined genotype of FAS-670GG/-1377AA conferred the highest risk of RPL. However, the FAS-670A/G variant had no impact per se on RPL risk, thus implying that gene-gene interaction is a significant factor to take into account for RPL susceptibility [21, 51]. No association with RPL was reported with FAS-1377G/A in our study. The differences in the RPL diagnostic criteria (≥ 3 pregnancy loss, ≤ 12 GWs) and in the genetic population background may account for these conflicting results. Further studies are needed to confirm whether the findings of our study apply to other definitions of RPL.

In agreement with a previous study with Brazilian and Asian populations [22, 27], our findings suggest that the FAS-670A/G SNP does not affect the risk of RPL in our population. Nevertheless, this SNP seems to play a role in other pregnancy disorders, such as preterm premature rupture of fetal membranes [52], preeclampsia [20, 23, 26], and fetal growth restriction [19]. Additionally, FAS-670A/G has been reported to play a role in nonrelated pregnancy conditions, such as cancer and autoimmune diseases [30, 44, 45, 53].

Regarding FAS-L variants, no association with RPL risk was observed. In contrast with our findings, Banzato et al. studied a Brazilian population and reported statistically significant differences in the FAS-L-844C/T genotype and allele distributions between the RPL and control patients. Additionally, among women with RPL, they observed higher levels of FAS-L mRNA expression, but no correlation with the FAS-L-844C/T genotypes was observed. Once again, conflicting results possibly reflect the RPL definition criteria adopted by different studies. Additionally, it is important to note that the RPL definition varies regarding the number of previous miscarriages among different international guidelines. Nevertheless, studies suggest that women who report two or more miscarriages should undergo clinical evaluation, as the risk factors and the chance of a miscarriage are similar to those with three or more miscarriages [54–56].

Our study is the first to evaluate the role of BAX-248G/A and BCL-2-938C/A SNPs in RPL susceptibility, and consequently, our study is the first description to suggest that protection is conferred by the BAX-248GA genotype in idiopathic primary RPL. Studies evaluating these SNPs in pregnancy disorders are scarce. However, the BAX-248G/A SNP has attracted much attention to malignant disorders [36, 40]. In this context, low levels of BAX expression in untreated CLL patients are associated with BAX-248GA/AA genotype carriers [39, 40]. Furthermore, the BAX-248GA genotype and the BAX-248A allele confer a decreased risk for systemic lupus erythematosus development [24]. A defective triggering of apoptosis characterizes both disorders previously cited (i.e., CLL and systemic lupus erythematosus). Although we were not able to investigate the association of BAX-248A/G SNP with the apoptosis index in endometrial biopsies, it is tempting to speculate that BAX-248GA heterozygotes have a better control of BAX expression and, thus, have a well-balanced proapoptotic and antiapoptotic ratio in comparison with the homozygous state. Nevertheless, the homozygous state of the BAX-248A allele was found only in women with primary RPL in our study. Additional studies are necessary to confirm our findings and to elucidate the functional relevance of BAX-248A/G on pregnancy outcome. Notably, the association of the BAX-248GA genotype with a decreased risk of primary RPL was significant even after the adjustment for potential confounding factors (ethnicity, smoking, and alcohol consumption). However, we highlight that external factors (i.e., psychological stress, endocrine disruptors, and air pollution) that were not covered in this study may influence women’s reproductive health and should also be further investigated [2, 57, 58]. The association of ethnicity with RPL susceptibility in black and Asian women has been reported [59, 60].

In the current study, we were not able to measure smoking and alcohol intake. We considered that even at a low dosage, alcohol consumption is a risk factor for adverse pregnancy outcomes, including miscarriage [2, 61]. Additionally, the observed effect of smoking on RPL susceptibility is paradoxical since smoking affects reproductive health, and the risk of miscarriage is dose-dependent [2, 62]. On the other hand, it has been suggested that smoking during the latter half of pregnancy may be protective against preeclampsia [63–65]. Thus, the association of smoking and the BAX-248GA genotype with RPL protection should be further investigated. Interestingly, a high expression of PGF, HIF1A, and the proapoptotic genes TP53 and BAX in villous samples from electively terminated, early pregnancies has been associated with maternal smoking [66]. In the same study, the authors speculate that maternal smoking influences placental morphological changes by inducing angiogenesis via PGF and apoptosis via the mitochondrial pathway (TP53 and BAX/BCL2) [66]. Since BAX expression is regulated by the TP53 protein [67], which is a transcription factor, it is likely that the effect of combined SNPs (epistasis) in both genes modifies apoptosis signaling and disease susceptibility. For example, the joint effect of the genotypes TP53, Arg72pro (rs1042522), and BAX-248AA is associated with the risk of head and neck squamous cell carcinoma [68]. Notably, a previous study from our group reported that genetic variation in the genes of the TP53 network influences RPL susceptibility, suggesting that impaired post-implantation mechanisms, such as p53-mediated embryo selection through apoptosis, contribute to recurrent pregnancy loss [69].

The distinction of RPL subclinical types is hardly made. Therefore, women with secondary RPL represent a poorly understood group and might represent a different entity of a characteristic phenotype. We observed an association of the BCL-2-938CA+CC genotypes with secondary RPL, although it lost its statistical significance after Bonferroni correction, and we highlight the need for further validation in more extensive studies. Previous studies that evaluated healthy endometrial tissues have reported that BCL-2 expression is high in the proliferative phase and gradually decreases toward the menstrual phase, with low or even absent expression after the middle secretory phase [70]. However, BCL-2 expression in the placenta is generally low throughout the gestational period, whereas BAX expression gradually increases by the end of pregnancy, thus implying that the regulation of both expression and the balance between BCL-2/BAX is spatiotemporal during pregnancy [71]. Therefore, imbalances between the proapoptotic and antiapoptotic proteins from the BCL-2 family might impair the delicate balance between programmed cell death and proliferation, thus increasing the risk for pregnancy disorders. In summary, data from the current study suggest that BAX-248AG is a protective genotype for primary RPL and that the BCL-2-938CC and C genotypes are risk factors for secondary RPL, but these findings will require confirmation in more extensive studies due to the low frequency of some genotypes.

In conclusion, our study is the first to evaluate the role of the main SNPs in both extrinsic and intrinsic apoptosis pathways in RPL susceptibility. Our data suggest that the regulatory variants in apoptosis-related genes contribute to the RPL risk in a Brazilian population. The association of the BAX-248G/A polymorphism with RPL risk implies that maternal genetic susceptibility for RPL has a potential contribution from genetic variants in regulatory regions that are involved in the delicate balance of endometrium cell turnover (cell death and proliferation). Therefore, apoptosis-related genes are potential candidates for future studies about healthy pregnancies and the spectrum of pregnancy disorders, as well as potential molecular biomarkers/targets for identifying women with fertility issues.

Funding information

This study was supported by Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (grant no. 14/2253-2 PqG) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (grant nos. 473115/2011-5 and 305839/2015-2).

Compliance with ethical standards

This study was approved by the Research Ethics Committee of the HCPA (HCPA-CEP-CPPG) under protocol number #11-242. We obtained written, informed consent according to the Declaration of Helsinki from all individual participants included in this study.

Conflict of interest

The authors declare that they have no conflict of interest.