Low-molecular-weight-heparin increases Th1- and Th17-associated chemokine levels during pregnancy in women with unexplained recurrent pregnancy loss: a randomised controlled trial

Low-molecular-weight heparin (LMWH) is widely used to treat recurrent pregnancy loss (RPL) because of its anti-coagulant effects. Although in vitro studies have suggested additional immunological effects, these are debated. We therefore investigated whether LMWH could modulate immune responses in vivo during pregnancy of women with unexplained RPL. A Swedish open multi-centre randomised controlled trial included 45 women treated with tinzaparin and 42 untreated women. Longitudinally collected plasma samples were obtained at gestational weeks (gw) 6, 18, 28 and 34 and analysed by multiplex bead technology for levels of 11 cytokines and chemokines, chosen to represent inflammation and T-helper subset-associated immunity. Mixed linear models test on LMWH-treated and untreated women showed differences during pregnancy of the Th1-associated chemokines CXCL10 (p = 0.01), CXCL11 (p < 0.001) and the Th17-associated chemokine CCL20 (p = 0.04), while CCL2, CCL17, CCL22, CXCL1, CXCL8, CXCL12, CXCL13 and IL-6 did not differ. Subsequent Student’s t-test showed significantly higher plasma levels of CXCL10 and CXCL11 in treated than untreated women at gw 28 and 34. The consistent increase in the two Th1-associated chemokines suggests a potential proinflammatory and unfavourable effect of LMWH treatment during later stages of pregnancy, when Th1 immunity is known to disrupt immunological tolerance.


Materials and Methods
An open multi-centre randomised controlled trial, registered at EudraCT (protocol number: 2010-022715- 19) was conducted in Sweden from 1 January 2012 to 31 December 2015, where pregnant women with a history of uRPL were included from six centres in the south of Sweden: Helsingborg, Lund, Kalmar, Jönköping, Karlskrona and Linköping. The primary outcome measure was immune modulation throughout pregnancy. Women with uRPL who were referred to one of the gynaecological clinics or a midwife centre were invited to participate in the study. eligible participants. Women who presented to the above units with uRPL, defined as three or more spontaneous consecutive miscarriages before 22 weeks of gestation without a known cause of RPL, were eligible for participation in the study. Only pregnant women were included and pregnancy was determined after confirming viable pregnancy by ultrasound. Inclusion and exclusion criteria (shown in supplementary Table I) were applied to those who accepted participation in the study, based on medical history and records. Participating women had previously undergone a standardised diagnostic workup according to national recommendations, including: a) collection of familial and personal medical, gynecological and obstetrical history with specific references to previous miscarriages; b) gynecological examination; c) transvaginal ultrasound; d) hysterosonography; e) endocrine evaluation panel: TSH, FT4; e) karyotype of both partners; f) immunity panel: anti-phospholipid antibodies, lupus anticoagulant, anti-cardiolipin antibodies, anti-β2GPI, anti-thyroid antibodies (anti-thyroid peroxidase and thyroid receptor antibodies); g) thrombophilia screening: protein C, protein S, homocysteine and determination of the following mutations: factor V Leiden, factor II prothrombin. This workup was aimed to identify proven causes of RPL. When all the above known causes for RPL had been excluded, women were diagnosed with uRPL and were included in this study. Written informed consent was obtained from all participants.
From the 129 women who were assessed for eligibility, 87 were included for randomisation: 45 women to treatment and 42 women to a control group. Charactaristics of included women is presented in Table 1. Another six patients were excluded during the study because of lack of compliance to the study protocol or fulfilling withdrawal criteria (Fig. 1, Supplementary Table 1), resulting in 81 patients completing the study to the end of their pregnancies. Eight women miscarried and 73 continued beyond 22 gestational weeks (gw), see Table 2. One woman with miscarriage and another three women with full-term live births had missing blood samples, and were therefore excluded from the analyses regarding cytokines and chemokines.
intervention. Once pregnancy was verified by ultrasound, all 87 eligible participants were recruited and randomised to one of the following groups.
Group 2 (control group) included 42 patients who neither received active treatment, nor placebo.
Randomisation. Enrolled patients were randomised through sequentially numbered closed envelopes opened consecutively at each inclusion point. Eligible patients were randomly assigned to one of the two study groups according to the envelope noting the intervention type. Once allocation was obtained, it could not be changed.
outcomes. The main outcome measures referred to the potential immunological effects of LMWH in vivo.
Clinical outcomes are shown as descriptive data including miscarriage rate, occurrence of pregnancy complications and any side effects of the drug used, either to the mothers or their infants.
www.nature.com/scientificreports www.nature.com/scientificreports/ Follow up schedule. All study participants followed the national guidelines regarding antenatal care with midwife check-ups during pregnancy and 8 weeks' postpartum, routine ultrasound screening at gw 18-20 and sometimes an additional chromosome (13,18,21) screening with combined ultrasound and blood sampling (PAPP-A, hCG). Additionally, participants underwent a transvaginal ultrasound at gw 10 and 12. Three other scans were done transabdominally around gw 18-20 (routine ultrasound according to national guidelines), 28 and 34 weeks to assess growth estimation, umbilical doppler measurement, cardiotocography (CTG) and questioning for adverse events at all check-ups. The antenatal records of the patients, including foetal viability, drug treatment side effects or any other complications, were recorded on the data collection sheet at each visit up to 2 weeks' post partum or miscarriage. In the LMWH-treated group hemoglobin, thrombocytes, APTt and PK-INR were controlled 2-4 weeks after starting LMWH following the local clinical routines. If no abnormal values in the above parameters were observed, no further hematological check-ups were done. plasma samples collection. Longitudinally collected blood samples were obtained from 87 women with uRPL enrolled in the current study assessing immunological effects of LMWH treatment. Samples were collected at four different time points during pregnancy: inclusion time point, before starting treatment (median gw 6, range 5-10), gw 18   32-37), and at 2 weeks' post partum. If a woman miscarried, the pregnancy stopped before the second sample time-point, and only inclusion samples were obtained. As we performed paired data analyses to evaluate LMWH effects during pregnancy, women with miscarriages were not included in the further analyses. Blood samples were collected in BD Vacutainer tubes, and after centrifugation at 1500 g, plasma was aliquoted and stored at −70 °C until use.

Data analysis and statistics.
A priori analysis showed that the sample size of 70 women was sufficient to detect a 40% decrease or increase in an immunological parameter (CXCL10) with a power of 80% and an alpha value of 0.05. Since the Kolmogorov-Smirnov test showed that data on cytokines and chemokines were not normally distributed, they were normalised by logarithmic transformation. First, Linear Mixed Models were used to evaluate differences between treated and untreated women in any of the cytokines and chemokines during the course of pregnancy. If p < 0.05, a post-hoc test was done to decide at which time point(s) and in which direction there was a difference. As only three chemokines were <0.05 in the linear mixed model test, the post-hoc testing in these cases was done with ordinary Student´s t-test or Fisher's exact tests; the latter was used in one case (CCL20) because of the low proportion of samples with detectable levels. Student´s t-test on log-transformed values was used to analyse differences in cytokine/chemokine levels at inclusion and at 2 weeks' post partum. Data were expressed as geometric mean and 95% confidence intervals in the figures and tables. P values < 0.05 were considered statistically significant. All data were analysed using SPSS version 24 (Armonk, NY: IMB Corp.) and graphs were made using GraphPad Prism version 7.0 (La Jolla, CA, USA). ethical approval. The   www.nature.com/scientificreports www.nature.com/scientificreports/

Results
Demographic, clinical and immunological data at inclusion. A total of 87 women participated: 45 in the treatment and 42 in the control group (Fig. 1). The two groups were similar in age, parity, number of previous miscarriages, gestational age, and body mass index at the time of enrolment (descriptive data are presented in Table 1). There was a similar frequency of miscarriage in treated and untreated women; 12% (5/42) and 7.7% (3/39) respectively.
Although 81 women without miscarriage completed the study, chemokines and cytokines were not measured in four women because of missing samples (Fig. 1). Of the remaining 77 women, chemokine levels were compared between women with an early miscarriage (n = 7) and women with a pregnancy beyond gw 22 (n = 70). CXCL11 was the only chemokine that differed between groups (Student's t-test on log-transformed data p = 0.013), showing significantly higher levels at inclusion, before treatment was started, in women who miscarried (geometric mean 49 pg/ml, 95% CI: 30-98) compared with the women who continued their pregnancy > gw 22 (geometric mean 27 pg/ml, 95% CI: 24-31, Supplementary Fig. 1).

Longitudinal comparisons of cytokine and chemokine plasma levels in LMWH-treated versus non-treated women.
A mixed models test based on all samples during pregnancy (inclusion, gw 18, 28 and 34) showed that LMWH-treated women differed compared with non-treated women in plasma levels of the Th1-associated chemokines CXCL10 (p = 0.007) and CXCL11 (p = 0.000) (Fig. 2a,b, Table 3.) A subsequent unpaired Student's t-test showed significantly higher levels of both these chemokines at gestational weeks 28 (CXCL10 p = 0.048, CXCL11 p = 0.023) and 34 (CXCL10 p = 0.026, CXCL11 p = 0.005) in treated compared with untreated women. The mixed models test also showed a significant difference (p = 0.037) between the treated and the untreated groups regarding the Th17-associated chemokine CCL20 (Fig. 2c). Since CCL20 levels were below the limit of detection in many samples, Fisher's exact test was used for post hoc testing, showing more samples with detectable levels in treated women (15% and 12% at gw 18 and 34 respectively) compared with untreated women (0% at both gw 18 and 34, p = 0.025 and p = 0.053 respectively). Except for CXCL10, CXCL11 and CCL20, linear mixed models did not show any other differences in the cytokine and chemokine levels throughout pregnancy when comparing the treated and untreated groups (Table 3), and therefore no further post-hoc testing www.nature.com/scientificreports www.nature.com/scientificreports/ was done. With the exception of CCL20, the analysed cytokines and chemokines were present at detectable levels in most samples (Supplementary Table 2). Cytokine and chemokine levels 2 weeks' post partum did not differ significantly between the treated and the untreated groups (data not shown).  www.nature.com/scientificreports www.nature.com/scientificreports/ Descriptive data of obstetrical outcomes. The live birth rate was high in both the treated (86%) and the untreated (90%) groups ( Table 2). The studied women with uRPL had a nominally higher rate of premature births, both in the study and the control group with 12% and 13%, respectively, compared with the background population in Sweden with a premature rate of 4% for live births 44 . The prevalence of intrauterine foetal deaths (IUFD) was nominally higher in the studied population (3%) compared with the background population where the incidence of IUFD is 3-4/1000 44 . The occurrence of preeclampsia, placental abruption, caesarean delivery or postpartum bleeding was similar in the treated and the untreated groups, as were neonatal parameters including mean gestational age (weeks), occurrence of small for gestational age and mean birthweight ( Table 2).

Adverse events.
No serious maternal adverse events were reported. There were small vaginal bleedings in four treated women and three untreated women, and heavy bleeding at childbirth in two untreated women. None of the LMWH-treated women had heparin-induced thrombocytopaenia and only one woman was reported to have pain or bruising at injection sites, although this could be underreported from patients as this information was not systematically requested. Regarding foetal malformations, one pleural effusion was found in the treated group and one gastroschisis in the untreated group. Intrauterine growth restriction was seen in three cases with treatment and none in the untreated group, while neonatal death occurred in one of the treated women and one IUFD occurred in each group.

Discussion
In this RCT of in vivo LMWH effects in pregnant women with uRPL, the main finding was a significant increase in plasma levels of the Th1-associated chemokines, CXCL10 and CXCL11, and possibly also of the Th17-associated chemokine, CCL20, in the second and third trimesters of pregnancy in women treated with LMWH compared to a control group without treatment. In contrast, LMWH did not decrease proinflammatory or increase anti-inflammatory/Th2-associated cytokines and chemokines. Thus, the only effects recorded would be supposedly unfavourable in relation to pregnancy, while other pregnancy-promoting effects not covered in the 11-plex panel may still occur.
To our knowledge, this is the first study assessing the immunological effects of LMWH in vivo in women with uRPL. Strengths in this study were the design as a randomised controlled study and the longitudinal sampling during the whole pregnancy. Also, we used a relevant panel of markers and not just one marker, and in general the analytes were measurable in plasma. A potential limitation was that peripheral blood does not represent the foetal-maternal interface environment. On the other hand, LMWH cannot pass over the placental barrier 45 and can therefore directly modify only the maternal side during placentation and during pregnancy. As it was an open study, the women and the healthcare providers were aware of that the women did or did not get the treatment, which could have influenced psychological mechanisms that in turn might affect inflammation 46 . Participation in a study suggests possible "tender, love and care" effects, which could also have psychological effects in a positive direction 47,48 . The clinical outcome, which showed a similar rate of pregnancy complications in the treated and non-treated group, support the notion of a strong "tender, love and care" effect. This effect might also be described by the higher pregnancy success rate than expected, based on the historical success rate in a similar population 47,48 The high pregnancy success rate could also be explained by the inclusion criteria that only viable pregnancies confirmed by ultrasound were included.
A consistent finding in our study was the increased plasma levels of the Th1-associated chemokines CXCL10 and CXCL11, which showed significantly higher levels in gestational weeks 28 and 34 in LMWH-treated compared with untreated women. CXCL10 and CXCL11 are both induced by IFN-γ and they bind to the same CXCR3 receptor on Th1 cells, which they both recruit 28,29,34 . Thus, their close relationship strengthens the biological relevance of the finding. A Th1 skewed immune response is known to affect pregnancy outcomes negatively in both murine models and in humans 8,23,25,43,[49][50][51][52][53] . Thus, the Th1-associated and potentially proinflammatory effects of LMWH that we observed in vivo are rather contradictory to the hypothesised anti-inflammatory and pregnancy-promoting effects that have been assigned to LMWH treatment. A Th1 skewing by LMWH is in line with our previous in vitro findings where LMWH enhanced production of the Th1 associated cytokine IFN-ɣ 24 , the cytokine that induces production of CXCL10 and CXCL11. Although LMWH treatment was associated with increased plasma levels of CXCL10 and CXCL11, we cannot draw any conclusion from the present study regarding the clinical effect of this finding.
Of note, a controlled Th1-associated proinflammatory setting is important in the early implantation phase (and during the window of implantation) 25,54,55 . This setting leads to the expression of genes implicated in endometrial receptivity, which is later followed by an anti-inflammatory response essential for post-implantation embryo survival and development, and for a balanced trophoblast invasion 56 . The early potential requirement of a Th1 response could imply that LMWH should rather be used at an earlier time point during pregnancy.
Although the role of Th17 in pregnancy is uncertain, it is known to reflect a mainly proinflammatory profile 39 . Our study showed that the Th17-associated chemokine CCL20 at gw 18 and 34 was more often above the detection limit in the women treated with LMWH compared with untreated women. The low detectability of CCL20 complicates the interpretation; still it is likely to be higher in the treated group. Higher levels of Th17-associated CCL20 would be unfavourable considering the reported increase in Th17-associated responses in non-pregnant women with a history of uRPL 43 . Th17 is also relevant in the context of a ratio to Treg cells since these subsets show a plasticity to switch to one another and since Treg cells are main inducers of tolerance during pregnancy [39][40][41][42] . Our earlier in vitro study showed that LMWH decreased Treg cells and increased Th17-associated CCL20 production 24 . In the present study, we did not evaluate markers of Treg cells since they do not have a specific profile in chemokine receptor expression. A previous in vivo study of women with thrombophilia showed that LMWH-treated women had higher levels of Treg cells compared to women without LMWH treatment during pregnancy 57 . However, Foxp3 was the only marker used to define Treg cells, which implies a risk of including www.nature.com/scientificreports www.nature.com/scientificreports/ activated non-suppressive Foxp3-expressing cells 58,59 . Also, the scenario of lowered Treg cells in thrombophilia may not necessary be applied in the context of uRPL. Still, it would be very relevant to evaluate further any LMWH-mediated effects on Treg cells in vivo.
The lack of significant differences between treated and untreated women regarding all other analytes investigated (n = 8; IL-6, CCL2, CCL22, CCL17, CXCL1, CXCL8, CXCL12 and CXCL13) indicates that LMWH does not have a major impact on several specific circulating cytokines and chemokine during pregnancy. Notably, Lissauer et al. 43 found Th1 and Th17-associated responses to be increased in non-pregnant women with a history of uRPL, i.e. the type of responses that we found to be increased by LMWH treatment in vivo.
Another finding of interest concerns the significantly higher plasma levels at inclusion (before treatment) of Th1-associated CXCL11 in women who later, regardless of treatment, miscarried compared to women who continued their pregnancy after gw 22. This finding is in line with several studies suggesting an enhanced systemic inflammatory response in women with uRPL, with higher levels of inflammatory cytokines, such as TNF, IFN-γ and IL-6, is a risk factor for miscarrying 49,50,[60][61][62][63][64] . Although a controlled proinflammatory profile is necessary for proper implantation and development of pregnancy, a too strong inflammatory response may be detrimental 23,54,55 .
Women with uRPL are often treated with empirical medical treatments without scientific evidence 4 . It is thus of great importance to study whether any treatment can help this subgroup to have a better prognosis. This study was not powered to answer if LMWH can improve any clinical outcome in women with uRPL, instead the focus was to assess any immunological changes induced by LMWH. As we found immunological differences not in favour of using LMWH as an anti-inflammatory drug during the second and third trimesters of pregnancy, our data do not support further RCTs to evaluate clinical outcomes using LMWH during these gestational weeks. Further studies are needed to explore the hypothesis of using LMWH-associated proinflammatory effects to ameliorate the defective implantation process in RPL in the earliest stages of pregnancy.

conclusion
The findings indicate a potential proinflammatory effect of LMWH treatment in vivo. The observed Th1 and possibly also Th17-like deviations do not support a beneficial immune effect of LMWH in vivo during the second and third trimesters of pregnancy.