Preeclampsia (PE) is a leading cause of perinatal maternal and neonatal morbidity and mortality, affecting 2–8 % of pregnancies in the industrialized world [1, 2]. Although PE places high burden on both the mother and the newborn infant, and despite considerable research over the past few decades, the precise cause and pathophysiology of PE remain largely unknown. The current leading hypothesis is that a defect in early placentation combines with impaired trophoblast differentiation, leading to multi-stage pathogenesis [3, 4].

The initial stage of the disease, which is largely asymptomatic, involves abnormal placentation, which leads to placental ischemia. This placental stage of PE is followed by widespread endothelial damage and dysfunction (known as the maternal stage) mediated by soluble factors from the placenta entering the maternal circulation. Given the high complexity of the disease, screening for PE and establishing a clear diagnosis of PE remain challenging in clinical practice. Nevertheless, the ability to identify patients at risk for PE and the ability to accurately diagnose patients with PE are essential in order to prevent PE-associated morbidity and mortality.

Cytokeratin 19 is a member of the keratin family of intermediate filament proteins responsible for maintaining the structural integrity of epithelial cells [5]. Cytokeratin 19 and other keratins are often used to differentiate cells of epithelial origin from hematopoietic cells in tests that evaluate circulating tumor cells in peripheral blood or trophoblastic particles in maternal circulation [6, 7]. When they occur in the peripheral circulation, cytokeratins are present primarily as partially degraded, individual protein fragments or complexes. Cytokeratin fragment 21–1 (CYFRA 21–1), a fragment produced from Cytokeratin 19, has been studied in patients with lung, esophageal, and gynecological cancers and has been shown to be a clinically useful prognostic marker [8–13]. Cytokeratins and their fragments are released into the circulation via several mechanisms, including cellular apoptosis, necrosis, abnormal mitosis, and/or spillover from proliferating cells, and these cellular mechanisms are also present in patients with PE [14–17].

Our group and others have reported increased levels of CYFRA 21–1 in the serum of preeclamptic women as well as in umbilical cord blood harvested from preeclamptic women [18–21]. Furthermore, serum CYFRA 21–1 levels have associated with disease severity and appear to be unaffected by antihypertensive medications [19, 20]. However, little is known regarding the dynamics of CYFRA 21–1 during pregnancy, regardless of whether the pregnancy is uneventful or complicated by PE. Given that the placental stage of PE precedes the maternal stage, we hypothesized that the pathogenic mechanisms of PE lead to increased levels of CYFRA 21–1, and this increase may precede the onset of PE-associated symptoms, thereby providing a marker that could be used to predict the risk of developing PE and/or to diagnose PE at an early stage.

To test this hypothesis, we measured CYFRA 21–1 levels in serum collected from women who had an uneventful pregnancy and from women who subsequently developed PE. Lastly, we examined the value of using CYFRA 21–1 to predict and/or diagnose PE.

The primary goal of this study was to measure the dynamics of maternal serum CYFRA 21–1 levels during uneventful pregnancy and during pregnancy complicated by PE. Our secondary goal was to determine the value of using maternal serum CYFRA 21–1 levels to diagnose PE and/or predict the risk of developing PE in asymptomatic women. Using samples obtained from the prospective longitudinal Biobank study, which includes sequential blood samples obtained at up to seven predetermined time points during pregnancy, we measured the time course of maternal serum CYFRA 21–1 levels. Our results revealed three major findings. First, the time course of maternal CYFRA 21–1 levels differs between uneventful pregnancies and pregnancies that are complicated by PE. Second, serum levels of CYFRA 21–1 are significantly higher in symptomatic women with PE (i.e., in our PE_state group) compared to women with an uneventful pregnancy, indicating that this circulating protein fragment is a promising biomarker for diagnosing PE. Lastly, the value of using maternal CYFRA 21–1 to predict the long-term occurrence of preeclampsia appears to be limited. However, our findings indicate a trend towards elevated maternal CYFRA 21–1 levels preceding the short-term occurrence of preeclampsia in asymptomatic women.

Worldwide, preeclampsia is estimated to account for approximately 50,000 cases of perinatal maternal death each year [2, 24]. Therefore, developing a reliable tool for identifying patients who are at risk for developing PE is an important first step towards preventing PE and PE-associated morbidity and mortality. However, the diverse clinical presentation of PE and the lack of a simple screen have hampered previous attempts to develop such a tool. Recently, biomarkers such as soluble fms-like tyrosine kinase 1 and placental growth factor have been used to diagnose and/or predict the short-term risk of PE in women with a clinical suspicion of PE [25–27]. However, the ability to identify women with a long-term risk for developing later in their pregnancy would constitute a significant advance in the field of obstetrics. A major obstacle against developing new strategies for diagnosing and predicting PE is our relatively limited understanding of the disease. For example, inadequate placental perfusion, hypoxia, and/or ischemia are all believed to play a role in the placental release of circulating soluble factors that can have widespread consequences, including maternal systemic endothelial damage and dysfunction, vasoconstriction, and end-organ ischemia, ultimately resulting in the clinical symptoms associated with PE [15, 28, 29]. The important role of widespread endothelial dysfunction in PE is underscored by the finding that applying serum obtained from preeclamptic women to in vitro cultured endothelial cells causes cytotoxicity [17, 30].

We previously studied the role of CYFRA 21–1 in PE and found significantly higher levels of CYFRA 21–1 in the serum of 32 preeclamptic women compared to 32 matched controls [20]. Our current results, which are based on a larger number of cases, support this previous finding. Here, we found that the serum levels of CYFRA 21–1 are approximately twofold higher in women with PE compared to women with an uneventful pregnancy. Intrigued by these findings, we hypothesized that the increased levels of CYFRA 21–1 in women with PE reflect endothelial damage and/or dysfunction due to circulating factors of placental origin. Given that the pathogenic changes responsible for the endothelial dysfunction clearly precede the resulting clinical symptoms associated with PE, we further hypothesized that changes in the serum concentration of CYFRA 21–1 also precede the clinical symptoms of PE. Interestingly, our analysis revealed lower levels of serum CYFRA 21–1 early in pregnancy (i.e., at visits 1 and 2, corresponding to gestational week 11–13 and 14–17, respectively) in women who subsequently developed PE compared to women whose pregnancy remained uneventful. Nevertheless, the dynamics of maternal serum CYFRA 21–1 levels differed between the asymptomatic women who later developed PE and the women who remained asymptomatic. For example, CYFRA 21–1 levels were higher in the PE_long group at visits 3 through 6; however, these differences were not statistically significant.

As an influence of maternal BMI and age on CYFRA 21–1 levels cannot be ruled out, we decided to calculate another mixed model, which additionally included the maternal variables pre-pregnancy BMI and age as predictors. The results of this analysis suggest a link between maternal BMI and maternal age, and levels of CYFRA 21–1 independently of future PE risk. The reasons underlying this link are unclear. One could hypothesize that maternal obesity is associated with endothelial dysfunction, abnormalities in placental angiogenesis, or both. Interestingly, our adjusted model revealed that the effect of elevated BMI led to a slight reduction of CYFRA 21–1 levels, which might be explained by an increased volume of distribution in women with elevated BMI. This issue, more specifically the effect of maternal BMI on serum angiogenic markers in pregnancy, was already adressed by other groups [31, 32]. In a longitudinal study, Zera et al. found associations between maternal BMI, sFlt-1, and PlGF in pregnancies affected by placental dysfunction and in normal pregnancies [33]. As the exact mechanisms underlying the association between maternal obesity and ischemic placental diseases like PE are still a matter of speculation, (i) the association of maternal characteristics like BMI and age with placental dysfunction and subsequent endothelial damage, (ii) the increased volume of distribution in obese women, and (iii) the influence of these characteristics on serum levels of biomarkers for PE deserve increased attention in future studies.

We also investigated the value of using maternal CYFRA 21–1 to predict the risk of developing PE later in pregnancy. With respect to the long-term predictive power of CYFRA 21–1, serum samples collected up until gestational week 22 (i.e., visits 1, 2, and 3) revealed no predictive power. Next, we analyzed samples obtained at later gestational periods (i.e., visits 4, 5, and 6) in order to evaluate the short-term predictive power of CYFRA 21–1. This analysis revealed that CYFRA 21–1 to some extent has short-term predictive power, as the AUC of the ROC curves for the increase in CYFRA 21–1 between two subsequent visits increased incrementally with increasing gestational age. We speculate that these findings may reflect the increasing degree of endothelial damage that occurs until the onset of visible symptoms associated with PE. Therefore, CYFRA 21–1 is an interesting candidate molecule for future studies investigating promising biomarkers for predicting the short-term occurrence of PE.

We acknowledge that the relatively limited number of patients in our study should be considered when interpreting the results. In particular, the small size of the PE_long group (i.e., ten patients) yielded relatively wide confidence intervals for the AUCs, thereby precluding our ability to draw definitive conclusions regarding the power of CYFRA 21–1 in predicting the subsequent onset of PE. However, given the low prevalence of PE in industrialized countries, it is extremely difficult to obtain large numbers of PE cases using a prospective longitudinal study design. Thus, in our study, only ten out of 461 patients who were initially recruited to the longitudinal arm of our Biobank ultimately met the criteria for being assigned to the PE_long group. Therefore, we included an additional study group comprised of symptomatic women with PE, thereby increasing the strength of our study.

Another potential limitation of our study is the relatively low number of samples obtained from PE cases at visit 7 (i.e., after gestational week 37). In clinical practice, the routine standard of care is to induce labor—or if warranted, deliver by cesarean section—in symptomatic women who present with either confirmed or suspected PE at gestation week 37 or later. Therefore, even our PE_state group (which contained 50 cases) included only a few samples from women with generally mild cases of PE at visit 7. This low number of samples late in gestation may explain at least part of the lack of difference in CYFRA 21–1 levels between preeclamptic women and control cases at term.

On the other hand, the prospective longitudinal design of the Biobank and the strictly predefined endpoints resulted in a well-characterized patient cohort that was monitored closely throughout the entire pregnancy. This robust study design enabled us to accurately measure the dynamics of maternal CYFRA 21–1 throughout pregnancy. In addition, because CYFRA 21–1 is also an established marker for various types of cancer, the reference values for CYFRA 21–1 that we obtained may help validate CYFRA 21–1 as a tumor marker in pregnant women.

Increased levels of serum CYFRA 21–1 in preeclamptic women may reflect endothelial damage and dysfunction due to circulating factors of placental origin. By comparing women whose pregnancy was complicated by PE with women whose pregnancy was uneventful, we found that CYFRA 21–1 is a promising biomarker for diagnosing PE. Although CYFRA 21–1 seems to have limited value for predicting the long-term occurrence of preeclampsia, our findings indicate that this marker may be useful for predicting the short-term occurrence of preeclampsia in asymptomatic women. Future prospective longitudinal studies will help determine whether this relatively simple, non-invasive test can be used to diagnose and/or predict preeclampsia.