Preeclampsia (PE) and eclampsia are the most severe hypertensive disorders of pregnancy, characterized by the occurrence of new-onset hypertension after 20 weeks’ gestation (Khedagi & Bello, Reference Khedagi and Bello2021). About 3−8% pregnant women in the United States are affected by PE, and this disease is also responsible for more than 60,000 maternal deaths and over 500,000 fetal deaths worldwide every year (Ma’ayeh & Costantine, Reference Ma’ayeh and Costantine2020). The complications of PE includes proteinuria, acute kidney injury, hepatic dysfunction, hemolysis and thrombocytopenia (Lambert et al., Reference Lambert, Brichant, Hartstein, Bonhomme and Dewandre2014). Unfortunately, PE is also the major risk factor for preterm birth and low birth weight, both of which are considered as the negative predictors of the child’s future health and cardiovascular risk (Ardissino et al., Reference Ardissino, Slob, Millar, Reddy, Lazzari, Patel, Ryan, Johnson, Gill and Ng2022).
Nowadays, although the advancement in therapies seems promising, delivery of the fetus in advance is the only definitive treatment strategy. Irrespective of the gestational age, severe PE patients with a gestational week ≥34 weeks or with unstable maternal or fetal conditions are recommended to commence delivery as soon as possible once the maternal condition is stable (Roberts et al., Reference Roberts, August, Bakris, Barton, Bernstein, Druzin, Gaiser, Granger, Jeyabalan, Johnson, Ananth Karumanchi, Lindheimer, Owens, Saade, Sibai, Spong, Tsigas, Joseph, O’Reilly and Ngaiza2013), while for severe PE patients with a gestational age of less than 34 weeks and with stable conditions, corticosteroids treatment are recommended to avoid fetal lung immaturity and to receive intensive care at a facility to maintain (Roberts et al., Reference Roberts, August, Bakris, Barton, Bernstein, Druzin, Gaiser, Granger, Jeyabalan, Johnson, Ananth Karumanchi, Lindheimer, Owens, Saade, Sibai, Spong, Tsigas, Joseph, O’Reilly and Ngaiza2013).
Early diagnosis and adequate human care are essential for PE patients. More importantly, more attention should be paid to fetal intrauterine status by daily monitoring of fetal movement and fetal heart changes. Despite this, there is an urgent need to develop effective drugs against the complications of PE.
Previous studies indicated that inflammatory status at the maternal-fetal interface throughout the pregnancy period is associated with PE progression (Perucci et al., Reference Perucci, Corrêa, Dusse, Gomes and Sousa2017; Raguema et al., Reference Raguema, Gannoun, Zitouni, Meddeb, Benletaifa, Lavoie, Almawi and Mahjoub2018; Subha et al., Reference Subha, Pal, Pal, Habeebullah, Adithan and Sridhar2016), suggesting the potential of inflammation-related genes serving as genetic biomarkers and drug targets for PE (Wang et al., Reference Wang, Li and Zhao2022). The serum levels of inflammation markers were found to associate with PE (Guven et al., Reference Guven, Coskun, Ertas, Aral, Zencirci and Oksuz2009). The study also found a significant difference in levels of high sensitivity C-reactive protein (CRP), interleukin-6 (IL-6) and tumor necrosis factor α (TNF-α) among uncomplicated pregnancies, and mild and severe PE patients (Guven et al., Reference Guven, Coskun, Ertas, Aral, Zencirci and Oksuz2009). Nevertheless, it was difficult to infer the causal relationship on the basis of observational evidence alone. Additionally, observational epidemiological studies are susceptible to confounding and reverse causation (Zheng et al., Reference Zheng, Baird, Borges, Bowden, Hemani, Haycock, Evans and Smith2017).
Mendelian randomization (MR) is an effective, potent and efficient method for determining the causal relationship between two correlated phenotypes (Greenland, Reference Greenland1990). MR, utilizing the summary statistics from genomewide association studies (GWAS), is a commonly used causal inference method to distinguish whether the exposure phenotype (e.g., inflammation) has a causal effect on the outcome phenotype (e.g., PE; Burgess et al., Reference Burgess, Dudbridge and Thompson2016; Dalbeth et al., Reference Dalbeth, Topless, Flynn, Cadzow, Bolland and Merriman2015) by using the genetic variants as instrumental variables (IVs).
In this study, single nucleotide polymorphisms (SNPs) strongly associated with inflammation biomarkers (e.g., CRP, IL-6, interleukin 1 receptor antagonist and TNF-α) were used as the IVs to infer the causal relationship between inflammation and PE. Since it was difficult to obtain the exposure and outcome data from the same person in a large cohort, a two-sample MR method was performed using the effect of IVs on the exposure and outcome phenotypes from two independent studies (Pierce & Burgess, Reference Pierce and Burgess2013).
Materials and Methods
GWAS Summary Data
PE heritability is estimated at 50−55%, with maternal genetic contribution risk of 30−35% and fetal genetic contribution risk of 20% (Gray et al., Reference Gray, Saxena and Karumanchi2018). The genetic association of PE was extracted from the analysis of the seventh release of the FinnGen consortium data (https://finngen.gitbook.io/documentation). The FinnGen consortium performed a large GWAS to identify the genetic variants associated with PE in 5265 cases and 160,670 controls of Finnish ancestry. The summary statistics identified 16,382,829 variants.
Based on the available GWAS summary data of inflammation biomarkers, the following inflammation biomarkers were selected as exposures: CRP, IL-1 receptor antagonist (IL-1ra), IL-6 and TNF-α. The genetic association estimates for the circulating levels of CRP were collected from a recent large meta-analysis of 88 GWAS studies performed on 204,402 European participants (Ligthart et al., Reference Ligthart, Vaez, Võsa, Stathopoulou, de Vries, Prins, Van der Most, Tanaka, Naderi, Rose, Wu, Karlsson, Barbalic, Lin, Pool, Zhu, Macé, Sidore, Trompet and Alizadeh2018). For IL-1ra, a GWAS of 90 circulating cardiovascular-related proteins conducted on 30,931 European individuals across 14 studies (Folkersen et al., Reference Folkersen, Gustafsson, Wang and Hansen2020) was used in the current study. The genetic association estimates for circulating serum IL-6 levels were extracted from a recent GWAS meta-analysis on 67,428 European participants across 47 cohorts (Ahluwalia et al., Reference Ahluwalia, Prins, Abdollahi, Armstrong, Aslibekyan, Bain, Jefferis, Baumert, Beekman, Ben-Shlomo, Bis, Mitchell, de Geus, Delgado, Marek, Eriksson, Kajantie, Kanoni, Kemp and Alizadeh2021). TNF-α association results were extracted from a meta-analysis of GWAS from 25 cohorts of 30,912 European participants (Prins, Reference Prins2016).
MR Analyses
The MR analysis has been well reported in a previous study (Davey Smith & Hemani, Reference Davey Smith and Hemani2014). To investigate the relationship between inflammation markers and PE, a two-sample MR method was used. The independent variants that are strongly associated with inflammation markers were selected as IVs (p < 5×10-8 and linkage disequilibrium (LD) r 2 < .01). Inverse-variance weighted method, weighted median and MR Egger (bootstrap) were used in this research. While the estimate for MR-Egger regression slope provides the pleiotropy-corrected causal effect, a published study confirmed that the weighted median approach affords some distinct superiorities over MR-Egger, such as its improved power of causal effect detection and lower type I error (Bowden et al., Reference Bowden, Davey Smith, Haycock and Burgess2016). In this research, to ensure more robust MR estimates, we also applied the weighted median method to complement the MR-Egger regression. The weighted median method may generate correct estimates even if up to 50% of SNPs are invalid IVs (Bowden et al., Reference Bowden, Davey Smith, Haycock and Burgess2016).
Sensitivity Analysis
Cochran’s Q test, a heterogeneity test, was performed to identify whether there was a higher heterogeneity among causal effects estimated using each variant individually rather than being expected by chance (Burgess et al., Reference Burgess, Dudbridge and Thompson2016).
To assess the possibility of other horizontal pleiotropic effects on how IVs affect inflammation biomarkers via other biological pathways, we performed a MR-Egger regression (Bowden et al., Reference Bowden, Davey Smith and Burgess2015). The intercept that deviated from the origin may provide evidence for potential pleiotropic effects across the IVs. All data analyses were conducted using R software (4.0.4 version).
Results
After harmonizing the exposure and outcome data and excluding the correlated SNPs, 25 SNPs for CRP, 4 SNPs for IL-1ra, 2 SNPs for IL-6 and 3 SNPs for TNF-α were used as IVs (Table 1, Supplementary Tables 1–3).
Note: SNP, single nucleotide polymorphisms; TNF-α, tumor necrosis factor α.
By applying the inverse-variance weighted method with fixed effects, CRP was found to have no significant association with increased PE risk (OR per 1-SD increase in CRP: 0.9, 95% CI [0.82, 1.04], p = .21). The risk of PE was not increased in patients with higher genetic predicted IL-1ra (OR per 1-SD increase in IL-1ra: 1.03, 95% CI [0.89, 1.20], p = .67). Patients with higher genetic predicted IL-6 had no association with increased PE risk (OR per 1-SD increase in IL-6: 0.86, 95% CI [0.55, 1.32], p = .48). However, the genetic predicted higher level of TNF-α was found to significantly increase the risk of PE (OR per 1-SD increase in TNF-α: 4.34, 95% CI [1.99, 9.45], p = .00022; Figure 1). The analysis using the weighted median method showed similar results with that using the MR Egger (bootstrap) method (Table 2).
Note: IVW, inverse-variance weighted; CRP, C-reactive protein; IL-1ra, interleukin 1 receptor antagonist; IL-6, interleukin-6; TNF-α, tumor necrosis factor α; NA, because there were only two SNPs used in weighted median and MR Egger method, no result was generated.
A heterogeneity test was applied to identify the possible effect of a higher heterogeneity among the estimated causal effects using the variants individually than being expected by chance. The result suggested that there was no heterogeneity (I2 = 41.4%, 95% CI [0.0%, 82.2], p = .18). The slope estimation of TNF-α level on PE using MR-Egger also showed no significant difference (beta = 0.14, 95% CI [-0.026, 0.31], p = .35).
Discussion
In this study, MR analysis was performed to detect whether inflammatory markers (CRP, IL-1ra, IL-6, TNF-α) were causally associated with PE. The results showed that the genetically elevated TNF-α level had putatively causal effect on the increased risk of PE and no causal relationship was identified between other inflammatory markers (i.e., CRP, IL-1ra and IL-6) and PE.
Delivery generally resolves symptomology, suggesting that the placenta plays a major role in the pathophysiology of PE (Rosser & Katz, Reference Rosser and Katz2013). TNF-α overproduced by the placenta in response to local ischemia and hypoxia contributed to the increased TNF-α level in the plasma (Conrad & Benyo, Reference Conrad and Benyo1997). Previous studies suggested that TNF-α released by the placenta to maternal circulation could elevate the expressions of proinflammatory cytokines secreted by endothelial cells, like IL-6, IL-8 and MCP-1 (Shaw et al., Reference Shaw, Tang, Schneider, Saljé, Hansson and Guller2016). Meanwhile, the activation and dysfunction of the endothelium is the main cause for PE (Roberts et al., Reference Roberts, Taylor, Musci, Rodgers, Hubel and McLaughlin1989). Elevated expression of TNF-α can induce the systemic acute phase response, which consequently stimulates the liver to synthesize CRP (Hansson, Reference Hansson2005). However, in this study, we only identified TNF-α causally increased the risk of PE, but not CRP and IL-6.
Maternal adipokines (i.e., IL-6, TNF-α, leptin and adiponectin) is responsible for the linkage of maternal nutritional status and adipose tissue metabolism with placental function (Reslan & Khalil, Reference Reslan and Khalil2010). A previous study suggested that the common indices of obesity posed a higher genetic risk of reproductive disorders, like PE (Venkatesh et al., Reference Venkatesh, Ferreira, Benonisdottir, Rahmioglu, Becker, Granne, Zondervan, Holmes, Lindgren and Wittemans2022). Although TNF-α serum level was higher in PE patients compared with healthy controls, obesity was not associated with serum TNF-α level in both PE and control groups (Founds et al., Reference Founds, Powers, Patrick, Ren, Harger, Markovic and Roberts2008). This result suggested that adipose tissue did not causally lead to PE via changing TNF-α level in the serum.
There are some limitations that need to be mentioned. First, only a limited number of SNPs were used as IVs to infer the causal relationship between inflammation biomarkers (e.g., CRP, IL-1ra, IL-6, and TNF-α) and PE. Previous studies indicated that the power for recognizing the causal relationship between complex traits by using genetic variants as IVs was enhanced with the increasing variance explained by these genetic variants (Brion et al., Reference Brion, Shakhbazov and Visscher2013; Freeman et al., Reference Freeman, Cowling and Schooling2013). So, when more variants were employed, the power could be increased and the results more credible. Second, the IVs were identified from cohorts of both sexes. However, the GWAS results of PE were generated from females. If sex-specific GWAS data of these inflammation biomarkers were available, the results would be more robust.
In summary, we identified that the genetically elevated TNF-α level had putatively causal effect on the increased risk of PE. This finding provided a therapeutic target for PE treatment, and pregnant women with a higher level of inflammation biomarkers, especially TNF-α, should receive more antenatal care.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/thg.2023.27.
Data availability
All data were uploaded as supplementary materials.
Acknowledgments
We want to acknowledge the participants and investigators of the FinnGen study.
Financial support
None.
Competing interests
None.