Peri-implantation cytokine profile differs between singleton and twin IVF pregnancies.

Study recruitment occurred between October 2017 and December 2018. The process is outlined in Figure 1 . Patient characteristics by singleton (n = 25) vs di-di twin pregnancy (n = 6) are presented in Table 1 . There were no significant differences in age, body mass index, history of live birth and miscarriage, fresh or frozen transfer, and race/ethnicity. As is intuitive, all di-di twin pregnancies resulted from transfer of two blastocyst embryos; 36% of singleton pregnancies resulted from transfer of two blastocyst stage embryos. Maternal serum CXCL10 concentrations from the blood draw performed 9–10 days after blastocyst ET were significantly higher in singleton pregnancies when compared to twins. In the other time periods, CXCL10 levels were not significantly different between singletons and twins. This is shown in Figure 2A and Table 2 . For overall trends, Cohen’s D does show significance, with a value of −0.59 for blood draws day 9–15, 95% CI −1.15 to −0.04. Maternal serum TNF-α concentrations were not significantly different between singletons and twins at any time point, as shown in Figure 2B and Table 2 . Cohen’s D was not significant for the difference in serum TNF-α between singleton and twin pregnancies from days 9–15 after ET, with Cohen’s D −0.29, and 95% CI −0.83 to 0.26. IL-10 levels were not significantly different between singletons and twins at all time points. This is shown in Figure 2C and Table 2 . Cohen’s D was not significant for the difference in serum IL-10 between singleton and twin pregnancies from days 9–15 after ET, with Cohen’s D −0.13, and 95% CI −0.66 to 0.4. It has been previously reported that IL-10 to TNF-α ratio is differentially expressed in pregnancies resulting in healthy term singletons vs pregnancy losses. 9 This ratio represents a balance of early pregnancy immunotolerant signaling and inflammatory signaling. We reviewed this ratio in singleton and twin pregnancies and found no significant differences, and this is shown in Figure 2D . Twin pregnancies did have lower IL-10 to TNF-α ratios throughout, which would indicate overall increased inflammation. Cohen’s D was not significant for the IL-10 to TNF-α ratio between singleton and twin pregnancies from days 9–15 after ET, with Cohen’s D −0.32 and 95% CI −0.85 to 0.22. As anticipated, maternal serum β-hCG levels in the early implantation period differed between singleton and di-di twin pregnancies, as shown in Table 2 ; β-hCG levels were significantly higher in di-di twins compared to singletons at all studied time points. Correlations between CXCL10, TNF-α, and IL-10 and correlations between cytokines and maternal serum β-hCG levels were also examined ( Figure 3 ). While associations of statistical significance were noted between individual studied cytokines ( Figure 3A – C ), maternal serum β-hCG levels did not demonstrate any correlation of statistical significance with CXCL10, TNF-α, or IL-10 levels at any of the studied time points in singletons or twins ( P > .05) ( Figure 3D – F ). We examined whether the peri-implantation cytokine profile of singleton clinical pregnancy that resulted from transfer of 2 embryos had a cytokine signal that was significantly different from a singleton resulting from single ET. This approach allowed examination of plausible contributions of an implanting but failing second embryo to the peri-implantation maternal cytokine profile of a singleton clinical implantation that resulted from the transfer of two embryos. As summarized in Table 3 , the concentration of all analytes tested was not significantly different between singleton clinical pregnancies that followed transfer of one vs two embryos.

Details of the cohort have been described in previous publications. 4 , 9 Inclusion criteria for this prospective cohort study were female patients, ages 18–45, undergoing fresh or frozen blastocyst (day 5) ET with a serum β-hCG level >50 mIU/mL on the first serum draw, which was timed between days 9–12 following ET. Exclusion criteria were patients with chronic autoimmune disease (such as lupus, thyroid autoimmunity, ulcerative colitis, or Crohn’s disease); diabetes and hypertension requiring medication; diagnosis of endometriosis confirmed by laparoscopy or endometriomas on imaging; or current illness (in general, we excluded patients with underlying inflammatory process). We also excluded patients with prior pregnancy losses, unless the products of conception were genetically tested and determined to be chromosomally abnormal. This study was approved by the Yale Institutional Review Board (#2000021607) and subjects provided verbal consent. Patients were offered participation at the time of ET. Protocols for fresh and frozen ET IVF cycle were as per standard of clinical care. For fresh ET cycles, controlled ovarian hyperstimulation regimen followed standard clinical practice with gonadotropin dose and choice of GnRH analogue individualized based on provider preference and patient characteristics. Transvaginal ultrasound-guided egg retrieval was performed approximately 35 hours following hCG trigger, decision for insemination vs intracytoplasmic sperm injection (ICSI) was based on clinical characteristics, and resulting embryos were cultured until day 5 of embryo development, followed by intrauterine transfer of one or two embryos under ultrasound guidance. Vaginal progesterone was used for luteal phase support and was continued until approximately 9 weeks of pregnancy. Frozen ET cycles utilized exogenous oral estrogen and intramuscular progesterone for endometrial preparation, as per standard of care. Embryo transfer of one or two fresh or thawed blastocyst stage embryos was performed under ultrasound guidance on the sixth day of progesterone exposure. As per standard clinical practice, serum β-hCG quantitative assessment was performed between days 9–12 after ET; serum level of >5 mIU/mL was taken to reflect evidence of implantation. In all patients with evidence of implantation, blood samples were collected approximately every 48 hours until β-hCG was ≥2500 mIU/mL, at which time a transvaginal ultrasound study was scheduled for confirmation of an intrauterine pregnancy. At each time point, blood sample was collected by venipuncture into two vacutainer tubes, and samples were left at room temperature for 60 minutes to allow for clotting, followed by centrifugation (Thermo Scientific Sorvall ST 16) at 1690 g for 10 minutes at room temperature. One vacutainer tube was used to test for β-hCG concentration. The β-hCG assay was carried out using a Cobas e411 analyzer (Roche Instruments). The assay used chemiluminescent technology in a sandwich design. The lower limit of detection was 0.1 mIU/mL, and the interassay coefficients of variance (CVs) were <6%. Serum from the second vacutainer tube was aliquoted into 1.5-mL polypropylene RNase- and DNase-free microcentrifuge tubes and stored at −80°C until cytokine assay. Transvaginal ultrasound studies (GE Healthcare Logiq P5, 7.5 MHz) were scheduled between 5 weeks and 4 days of gestation to 6 weeks and 5 days of gestation (or, 20–28 days after blastocyst ET). Diagnosis of a singleton clinical pregnancy was based on ultrasound documentation of a normal intrauterine gestational sac with visible yolk sac, fetal pole, and presence of fetal cardiac activity. 14 A diagnosis of a di-di twin gestation was based on ultrasound documentation of two intrauterine gestational sacs with a lambda sign, each with a yolk sac, fetal pole, and fetal cardiac activity following ET of two blastocyst stage embryos. 14 Pregnancies were followed by periodic review of electronic medical records until delivery. A normal course of pregnancy (absence of pregnancy complications such as pre-eclampsia, gestational diabetes, abruption, or fetal growth restriction) and delivery of a liveborn at term (≥37 weeks) were pre-specified criteria for selection of serum samples for cytokine assays for this study, given our goal to generate normative data. Twin pregnancies that spontaneously reduced to a singleton gestation were not included in this analysis. Assays were performed using the SimplePlex ™ platform (ProteinSimple), an automated cartridge-based immunoassay with microfluidic technology (Aldo 2016). The cartridge is premade by the manufacturer according to individual needs with one to four analytes. Within the cartridge, all the steps of a sandwich ELISA are executed automatically, thus eliminating potential for human error, and allowing for fast (<80 minutes) and sensitive results. Samples are separated into parallel channels if analyzing multiple analytes, and each analyte is analyzed in triplicate glass nanoreactors, eliminating cross-reactivity when multiplexing and generating a triplicate reading of each analyte. Sample concentrations (pg/mL) are automatically calculated by fitting relative fluorescence units (RFUs) to calibration curve parameters provided with each cartridge and adjusted for possible user-defined dilution factors. The SimplePlex ™ platform has previously been characterized and validated by our group. 15 Cryopreserved serum samples were thawed and diluted 1:2 with diluent provided by the manufacturer; 50 μL of diluted sample was loaded into the cartridge. Data were exported to Microsoft Excel for analyses. The lower limits of detection were 0.49 pg/mL for CXCL10, 1.14 pg/mL for TNF-α, and 0.46 pg/mL for IL-10. The intra- and interassay CVs were 2.14%–2.89% and 2.55% for CXCL10, 3.3%–10.4% and 6.9% for TNF-α, and 2.1%–4.9% and 4.3% for IL-10. Data were examined for any outliers as means of the triplicate reads. Data distribution was examined, and for skewed data (CXCL10 and β-hCG), normal distribution was attained by log transformation of β-hCG and the reciprocal square root of CXCL10. Differences in serum levels of CXCL10, TNF-α, IL-10, and β-hCG between singletons and di-di twins by gestation (in days) were compared using Student’s t tests. We then estimated the difference of the individual cytokines between singletons and twins using Cohen’s D , an effect size measure. A positive Cohen’s D indicates that twins have higher cytokine levels than the singletons, while a negative Cohen’s D indicates the twins have lower levels than the singletons. Unlike P -value, the sample size has no influence on the statistical significance of Cohen’s D . 16 Additional analyses examined whether serum levels of analytes differed between singleton pregnancies resulting from transfer of a single blastocyst embryo (100% implantation rate) from those with 50% implantation rate—that is singleton pregnancy following transfer of two blastocysts, and we further examined whether the number of embryos transferred was relevant to the observed differences in the serum analytes between singleton and twin pregnancies. Correlations between the various analytes were examined using Pearson’s and Spearman’s correlation coefficients for data demonstrating Gaussian and skewed distributions, respectively. Statistical analyses and graphs were done in R software 3.5.1 (R Foundation) and Stata 15.1 (StataCorp LLC). Data are presented as mean ± standard deviation (SD) or median (range). A two-sided P -value <.05 and Cohen’s D 95% confidence interval (95% CI) ranges that did not cross “zero” were considered statistically significant.

We describe CXCL10, TNF-α, and IL-10 in the peri-implantation period of uncomplicated singleton and di-di twin pregnancies conceived by IVF. The novelty of our findings lies in this being the first report of serial changes in CXCL10 levels in the maternal serum in the peri-implantation period in di-di twin pregnancies and the first documentation of a distinctive pro-inflammatory cytokine profile in singleton compared to twin pregnancies. Maternal serum TNF-α levels in early healthy singleton pregnancies show that after an initial pro-inflammatory picture with average values around 6 pg/mL, TNF-α values decrease by the end of the first trimester to 5.5 pg/mL. 9 , 10 , 17 This study reconfirms the expected TNF-α concentration in the implantation phase in singleton pregnancies and adds information about twin pregnancies, showing lower initial TNF-α concentrations. IL-10 maternal serum values in singleton pregnancies have been shown to increase around implantation, likely guiding an immunotolerant response to the invading embryo. 9 , 18 , 19 This study shows that twins do not have exaggerated IL-10 production in the serum. The IL-10 to TNF-α ratio is also not different between singletons and twins, suggesting that, in general, the balance of inflammation and immunotolerance required to establish di-di twin pregnancy is similar to that of singleton pregnancies. The correlation between IL-10 and TNF-α is positive throughout this time frame, which also reinforces this concept of balancing between immunotolerance and inflammation. The correlation between IL-10 and 1/[square root (CXCL10)] is negative, which suggests that immunotolerant IL-10 may balance the inflammatory CXCL10 signal as well. CXCL10 binds to the CXCR3 cell surface receptor, which has been identified on T lymphocytes, natural killer (NK) cells, inflammatory dendritic cells, macrophages, B cells, and human blastocyst trophectoderm cells. 3 , 7 , 20 CXCL10 is therefore believed to be involved in signaling and recruitment of these cells. Unlike TNF-α, CXCL10’s serum concentrations in early human pregnancy were not well-established. In reproduction, CXCL10 is believed to play a key role in apposition of the blastocyst and adhesion of the trophectoderm to the endometrium. 4 , 7 , 21 – 25 CXCL10 secretion by endometrial cells is increased on day 14 of the menstrual cycle, and CXCL10 has been detected in endometrial aspirations prior to ET in IVF, 5 , 26 but secretion decreases by day 21 in humans who were not pregnant, 3 and also decreases in decidualized endometrium. 4 CXCL10 endometrial secretion, therefore, increases around the time of ovulation and fertilization, but decreases by the time of implantation, possibly to modulate its roles recruiting CD8+ T cells, promoting inflammation, and blocking angiogenesis. 4 , 20 , 27 – 29 While CXCL10 may be important at the time of blastocyst-decidual apposition, it may be detrimental to the progression of implantation. Our study furthers this research by identifying that in the early stages of implantation, approximately cycle day 30–32, the maternal serum profile of CXCL10 mimics the low levels described at day 21 in the endometrial secretion studies. A previous study showed a lowering of maternal serum levels of CXCL10 at 10-week gestation (162.7 pg/mL) compared to pre-pregnancy serum levels (around 400 pg/mL). 30 Our study shows that proximate to implantation, maternal CXCL10 serum levels are even further decreased in singletons and di-di twin pregnancies. The lower maternal circulating levels of CXCL10 in twins compared to singleton implantations may be related to a dose-dependent β-hCG-induced suppression of the CXCL10 promoter, as has been previously suggested. 4 Although we did not observe a correlation of statistical significance between serum β-hCG and CXCL10 levels, the magnitude of association as well as directionality of the relationship differed between singleton and twins; Spearman’s rho was 0.13 for days 9–11, 0.37 for days 12–13, and 0.09 for days 14–15 in singleton pregnancies ( P > .05 for all time points) compared to −0.6 for days 9–11, −0.54 days for 12–13, and −0.8 for days 14–15 in twins ( P > .05 for all time points). Despite the lacking statistical significance, the greater magnitude of association (ie, a more robust Spearman’s rho) and the inverse directionality of the relationship between β-hCG and CXCL10 observed in di-di twin pregnancies is intriguing and is in line with the hypothesized suppressive effect of β-hCG on CXCL10. 4 A larger sample size is needed to allow clarity if indeed higher maternal serum β-hCG levels in twins compared to singleton pregnancies may underlie the lower CXCL10 levels observed in twins compared to singleton pregnancies. Despite the novelty of our findings, our study has limitations in addition to the small sample size which merit acknowledgment. We do not have information on the pre-implantation levels of the studied cytokines and hence cannot comment on the magnitude of shift in serum levels of the studied cytokines at the earliest stage of implantation, with rise in serum β hCG level above 5 mIU/ml. Because our study population was limited to twin implantations that progressed to twin delivery, we are unable to examine whether peri-implantation cytokine profiles may vary in those twin implantations that undergo spontaneous reduction to singleton pregnancy. Because of the small sample size, our study was underpowered to detect differences between singletons and twins. Using the differences we found, we suspect a sample size of approximately 100 singletons and 50 twins would provide 80% power to detect an α of 0.05. We have characterized the early first-trimester maternal serum profiles of CXCL10, TNF-α, and IL-10 longitudinally in singleton and di-di twin pregnancies in humans. We have additionally examined whether and how serum levels of CXCL10 relate to the rise in β-hCG levels over a course of 2 weeks following implantation. Our findings provide novel insight into how CXCL10 levels differ between singleton and di-di twin pregnancies. Physiologically, it is plausible that there is some cross-talk between the invading embryo(s) and the endometrial stroma and decidual NK cells that are secreting the CXCL10. This could be regulated by β-hCG, which is higher in twin pregnancies, or by some other unidentified signal. Our study design does not allow us to examine the mechanisms that can explain the observed differences in CXCL10 levels between singleton and twin implantations. It is, however, reasonable to posit that given the known anti-angiogenic properties, the lower CXC10 levels in twins create an environment that is conducive to enhanced angiogenesis that is required for the establishment of dual placentation in twin pregnancies. Comparing biomarkers in singletons and twin gestations can allow a better understanding of targets that are of pathological relevance. Our overarching goal is to contribute to an improved understanding of processes that allow for normal and uneventful progression, and timely culmination of pregnancy, and our current work represents a miniscule step toward the ultimate goal. It is only when we have established normative data for healthy pregnancies (singleton and multiple) that can we meaningfully interpret and target aberrant profiles utilizing immune modulatory strategies to improve prognosis of at-risk pregnancies.

Implantation is characterized by an initial inflammatory reaction to the invading semi-allogenic embryo, which is then modulated to allow fetal development, growth, and eventually successful timely culmination of an uneventful pregnancy. Cytokines and chemokines are secreted signals that act on maternal decidual cells and on the blastocyst itself to play key regulatory roles in the processes of embryo apposition, trophectoderm adhesion to the uterine surface epithelium, and invasion of the decidua. 1 – 3 C-X-C motif chemokine ligand 10 (CXCL10) and tumor necrosis factor-alpha (TNF-α) are both pro-inflammatory cytokines known to be secreted by the human endometrium prior to implantation. 3 , 4 In patients undergoing in vitro fertilization (IVF), TNF-α and CXCL10 are both more highly concentrated in endometrial aspirations prior to embryo transfer (ET) in patients who achieve implantation compared to those who do not. 5 , 6 TNF-α is known to increase the transcription of other pro-inflammatory cytokines and recruitment of immune cells to the endometrium. 6 CXCL10’s receptor, CXCR3, is known to be on human trophectoderm, 3 , 7 and CXCL10 appears to contribute to dedifferentiation of vascular smooth muscle cells in the spiral arterioles, which inhibits vascular growth. 8 We have previously reported on the existence of a pro-inflammatory maternal milieu at the earliest stages of implantation, as reflected in higher serum levels of TNF-α at approximate gestation of 4 weeks. 9 A progressive shift toward an anti-inflammatory or immune tolerant environment was observed as healthy pregnancies progress in the first trimester, reflected in declining serum TNF-α and increasing IL-10 levels with advancing gestation, 9 and decreasing CXCL10/ β-human chorionic gonadotropin (β-hCG) ratio in early gestation. 4 Produced by the invading syncytiotrophoblast, β-hCG is known to suppress the CXCL10 promoter through histone methylation, thus reducing CXCL10 expression in the decidua. 4 A failing pregnancy has a cytokine expression profile consistent with heightened inflammation lasting for too long. 2 , 9 , 10 Our collective understanding on whether maternal immune responses and circulating cytokine profiles differ between singleton and multiple implantations is lacking. It is unclear whether successful di-chorionic di-amniotic (di-di) twin pregnancies induce a more extreme inflammatory response or more immunotolerance in the peri-implantation period compared to singleton implantation. Fisch et al have previously demonstrated that transferring increasing numbers of embryos expressing human leukocyte antigen G (HLA-G), an immunotolerant major histocompatibility complex, was more likely to result in multifetal gestation than when transferring increasing numbers of embryos without HLA-G, possibly indicating a need for increased immunotolerance in twin gestations. 11 More information is needed on the actual maternal response to two implanting embryos, as opposed to one. Our group is interested in examining the normal elements of the maternal immune system’s cross-talk between the embryo and endometrium at the earliest stages of implantation. By serial assessments of maternal circulating cytokine levels, we hope to glean “real-time” information on how easily accessible biomarkers may be utilized to reflect the rapidly changing maternal-fetal interface in the peri-implantation period. Our overarching goal is to elucidate any missteps in immune signaling sequences that may be relevant to occurrence of spontaneous miscarriage and recurrent pregnancy loss, 12 , 13 such that pregnancy prognosis may be improved through a targeted immune modulatory intervention. We herein have examined whether maternal responsivity to implantation differs between singleton and twin pregnancies resulting from IVF-ET. We hypothesized that the maternal circulating immune profile, specifically differences in pro-inflammatory and immunotolerant serum cytokine expression, will differ between singleton and twin pregnancies in the early implantation period.