Circulating Vascular Endothelial Growth Factor and Soluble fms-Like Tyrosine Kinase-1 as Biomarkers for Endometrial Remodeling Across the Menstrual Cycle

BioCycle was a prospective cohort study originally designed to determine the association of oxidative stress and endogenous reproductive hormone levels across the menstrual cycle in premenopausal women 23 . From 2005 to 2007, 259 healthy women were enrolled from western New York and prospectively followed for 2 menstrual cycles. All participants provided written informed consent and the University at Buffalo Health Sciences Institutional Review Board approved the study and served as the IRB under a reliance agreement with the National Institutes of Health. Of the 259 women (509 cycles) included in the original BioCycle study, one menstrual cycle from 96 women with confirmed ovulation were randomly selected through a computer generated algorithm for inclusion in the present ancillary study. Ovulation was confirmed using serum luteal progesterone measurements 24 , with cycle visits timed using fertility monitors 25 (Clearblue ® Easy Fertility Monitor, Iverness Medical, Waltham, MA, USA). Full cohort details, including detailed inclusion and exclusion criteria are available elsewhere 23 . At the screening visit, a questionnaire was administered to all women to determine eligibility. Briefly, women from age 18 to 44 years were enrolled if they self-reported their cycle lengths between 21 and 35 days. Women were excluded from the study if they had used long acting reversible contraception over the past 12 months or an oral contraceptive or other hormonal contraceptive over the prior 3 months. Women who reported trying to conceive, had been pregnant or breastfeeding in the preceding 6 months also were excluded. Other exclusion criteria included a history of infertility, laparoscopy confirmed endometriosis, polycystic ovarian syndrome (PCOS), fibroids, chlamydia, treatment for an infectious disease in the past 6 months, any antibiotic use in the past 3 months, untreated genitourinary infections in the past 6 months, and a self-reported body mass index (BMI) 35 kg/m 2 at screening. At baseline, participants completed questionnaires regarding menstrual, reproductive, and medical history, as well as demographic and lifestyle factors. Recruitment was inclusive with respect to race/ethnicity, which was self-reported using provided categories or through an open-text option. Collection of race/ethnicity data is required for research funded by the National Institutes of Health. Once enrolled , all women had height and weight measured using standardized protocols to calculate BMI. To time visits at specific cycle phases, fertility monitors (Clearblue® Easy Fertility Monitor, Iverness Medical, Waltham, MA, USA) and personal cycle length histories were used 25 , 26 . Each participant provided fasting blood and urine specimens at eight specific phases of the menstrual cycle, including menstruation, mid-follicular phase and late follicular phase, LH surge, estimated day of ovulation, and early, mid- and late luteal phase. Blood and urine samples were centrifuged for 10 minutes at 4°C and placed in a −80°C freezer. The total time between phlebotomy and freezing samples was less than 90 minutes. Concentrations of VEGF and sFLT-1 were determined in each of the eight menstrual cycle phases listed above in plasma, serum, and urine. VEGF and sFLT-1 were measured with quantitative sandwich enzyme immunoassay (EIA) (R&D Systems, Minneapolis, MN). The lower limit of detection (LOD) of VEGF was 9.0 pg/mL. For VEGF, the inter-assay laboratory coefficients of variation (CVs) were <11.2% for plasma, <8.0% for serum, and <12.6% for urine. The LOD of sFLT-1 was 8.46 pg/mL. For sFLT-1, the inter-assay laboratory CVs were <7.5% for plasma, <5.9% for serum, and <12.3% for urine. All instrument-derived values were recorded for measures below the LOD and were used in the analysis 27 . For VEGF, values below the LOD occurred in none of the serum samples, 0.1% of plasma samples (n=1), and 0.9% of urine samples (n=7). For sFLT-1, values below the LOD occurred in none of the serum or plasma samples, but 16% of urine samples (n=105). Eight measurements across the cycle were available for analysis in plasma in 100% of participants; in serum for 97% and 88% of participants for VEGF and sFLT-1, respectively; and in urine for 96% and 93% of participants for VEGF and sFLT-1, respectively. The reproductive hormones, estradiol, progesterone, FSH, and LH were measured in serum in each of the eight menstrual cycle phases with the use of solid-phase competitive chemiluminescent enzymatic immunoassays (Specialty Laboratories, Inc., Valencia, CA) on the DPC Immulite 2000 analyzer (Siemens Medical Solutions Diagnostics, Deerfield, IL). The CV for these tests reported by the laboratory were <10% for estradiol, <5% for LH and FSH, and <14% for progesterone.Averages of the eight VEGF and sFLT-1 measurements across the menstrual cycle were used to construct tertiles for each analyte within each biospecimen type (plasma, serum, urine). Demographic characteristics were compared by tertiles of VEGF and sFLT-1 in each medium. Chi-square and ANOVA tests were used to test for differences as appropriate. Descriptive statistics were used within each tertile for each analyte and biospecimen type to characterize the distribution of VEGF and sFLT-1. Pearson correlation coefficients were used to examine concentrations of VEGF and sFLT-1 determined in each biospecimen type (plasma, serum, and urine) at each menstrual cycle phase, and median concentrations were compared at each point in the menstrual cycle. Reproductive hormone concentrations were log-transformed to approximate normality for analysis. Harmonic models were used to evaluate associations of VEGF and sFLT-1 with hormonal patterns of estradiol, progesterone, LH, and FSH over the menstrual cycle, including assessment of changes in the mean levels over the menstrual cycle, amplitude of the change in levels from the nadir to the peak, and shifts of the timing of the peaks. Adjustments for BMI and age were made due to their association with hormones 28 – 30 . We also controlled for race as we found it was associated with VEGF and sFLT-1 in our data. Statistical analysis was conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC) and R software (Version 3.5, https://www.r-project.org/ ) for harmonic models.

Mean (standard deviation) age and BMI of participants were 28.7 (8.0) years and 24.5 (3.9) kg/m 2 , respectively ( Table 1 ). The mean age of menarche was 12.5 years old and 70.5% of participants were nulliparous. The majority of participants were White (52%). Fifty-six percent had a college diploma or more advanced degree. Only 2% were current smokers and 7.5% consumed at least 3 alcoholic beverages per week. There were no statistical differences (p> 0.05) in baseline characteristics and demographics when comparing participants by tertile of VEGF ( Table 1 ) or sFLT-1 ( Appendix 1 , available online at http://links.lww.com/xxx ) in serum, plasma, and urine. Median (25 th percentile, 75 th percentile) concentrations of VEGF across the menstrual cycle were 31.2 pg/mL (24.1, 56.9) in plasma, 194.1 pg/mL (125.4, 350.2) in serum, and 101.7 pg/mL (64.2, 165.8) in urine ( Table 2 ). Measurements of VEGF in plasma and serum were correlated at each menstrual cycle phase ( ρ ^ = 0.34 to 0.76 ) , while measurements in urine were uncorrelated with those in plasma or serum ( ρ ^ = − 0.18 to 0.12 ) ( Table 3 ). No variation in VEGF was detected over the menstrual cycle in any of the three biospecimen types ( Figure 1a ). No consistent associations were observed between VEGF and hormones ( Appendix 2 , available online at http://links.lww.com/xxx ), though we did observe associations between plasma VEGF and amplitude of estradiol ( Figure 2a ), and serum VEGF with phase shift of progesterone and FSH. Median (25 th percentile, 75 th percentile) concentrations of sFLT-1 across the menstrual cycle were 347.1 pg/mL (315.3, 386.3) in plasma, 421.9 pg/mL (383.9, 471.2) in serum, and 27.5 pg/mL (18.2, 46.5) in urine ( Table 2 ). Measurements in plasma and serum were correlated at each menstrual cycle phase ( ρ ^ = 0.39 to 0.68 ) , while measurements in urine were uncorrelated with those in either plasma or serum ( ρ ^ = − 0.13 to 0.20 ) ( Table 3 ). No variation in sFLT-1 was detected over the menstrual cycle ( Figure 1b ). Plasma sFLT-1 was associated with higher mean estradiol (β: 0.29, 95% CI 0.04, 0.54) and amplitude (β: 0.33, 95% CI 0.01, 0.65) per unit (pg/mL) increase in sFLT-1 ( Figure 2b ) ( Appendix 3 , available online at http://links.lww.com/xxx ). However, no other relationships between sFLT-1 and menstrual cycle hormones were detected.

In this prospective cohort of healthy premenopausal women, there were no changes in VEGF or sFLT-1 concentrations measured using plasma, serum, or urine across the menstrual cycle. VEGF was not associated with any reproductive hormone pattern consistently, although higher plasma sFLT-1 was associated with higher mean and amplitude of estradiol concentrations. Based on these findings, circulating VEGF and sFLT-1 are unlikely to be informative peripheral biomarkers for endometrial remodeling across the menstrual cycle, despite the potential role of these markers at the tissue level. Our findings of no cyclic changes in VEGF across the menstrual cycle are consistent with some 20 , 21 , 31 , but not all 17 , 19 , 32 prior investigations. Previous studies that have observed cyclic variation in VEGF differ from our study in important ways. First, one prior study included women with polycystic-appearing ovaries that have been associated with higher serum VEGF concentrations and may be responsible for the changes across the cycle observed in that study 18 . Second, another study observed cyclic changes among women undergoing controlled ovarian hyperstimulation in comparison to a control group of unmedicated ovulatory women 31 . The increase in serum VEGF in that study correlated with peak estradiol levels and it is postulated that the changes in VEGF in stimulated cycles is due to multi-follicular growth, corpus luteal function, and endothelial stimulation. The supraphysiologic levels of estradiol and multi-follicular growth seen in stimulated cycles may thus result in higher serum VEGF that may otherwise be negligible in the physiological menstrual cycle. Third, one study included older participants with mean age of 36 years (range 20-42 years) in contrast to our study, with mean age of 28.7 years (range 18-44) 17 . Overall, our work in a large healthy population of women selected to have regular menstrual cycles with carefully timed biospecimen collection is consistent with studies evaluating healthy regularly menstruating women which observe no cyclic changes in VEGF 20 , 21 , 31 , as opposed to pathologic states or stimulated cycles, and also extends prior work by evaluating its receptor sFLT-1. Since VEGF is produced throughout the body, there may be differences in VEGF measurement by biospecimen type, which has varied in prior studies. Serum VEGF reflects both systemic and locally derived VEGF 21 , 33 , 34 , as VEGF is secreted from platelets into serum during the clotting process after collection 33 and, subsequently, serum VEGF correlates with platelet number 35 . As plasma is void of platelets, it likely best reflects in vivo VEGF. Indeed, in our study, the median concentrations of VEGF over the menstrual cycle were six times higher in serum (194.1 pg/mL) than in plasma (31.2 pg/mL) in plasma. This is similar to another study which reported mean plasma VEGF of 5.8 pg/mL and a mean serum VEGF of 262 pg/mL in healthy menstruating women, noting, however, that most of the plasma concentrations measured in their study were below the cited range of detection 21 . Therefore, studies that measure only serum VEGF should be interpreted cautiously. In fact, all studies that reported menstrual cycle variation in VEGF used only serum analysis 17 – 19 . Despite the 6-fold difference in magnitude of VEGF concentrations, there were significant correlations between plasma and serum VEGF concentrations in our population though the strength of the correlation varied somewhat over the menstrual cycle. Similarly, in a study measuring circulating VEGF in serum and plasma in patients undergoing COH for IVF, serum and plasma VEGF were also significantly correlated 34 . Thus, plasma and serum correlations suggest that systemic VEGF is captured in both biologic specimens, although higher serum levels reflect platelet release of VEGF associated with blood collection and processing. In comparison, urinary VEGF was not correlated with either serum or plasma VEGF. This may due to the production of VEGF by the kidney and therefore may not reflect systemic levels 36 . For example, urinary VEGF has some predictive value for the development of preeclampsia, a disease that commonly includes renal dysfunction 37 . Though no cyclical changes in sFLT-1 were observed throughout the menstrual cycle, we observed associations with higher mean and amplitude of plasma estradiol. It is possible that estradiol acts as a hormonal stimulus to promote upregulation of the sFLT-1, an antagonist of VEGF, and our findings are supported by a study that demonstrated higher expression of sFLT-1 in endometrial biopsies collected during mid- and late proliferative phases 16 . Further work is needed to investigate regulation of angiogenesis as an inverse relationship was previously reported between increasing serum sFLT-1 and decreasing serum VEGF in pregnant women who develop preeclampsia 8 . As such, although changes in sFLT-1 were not noted during the menstrual cycle, it is possible that these associations may be spurious. Further, there may be utility in further study of VEGF and sFLT-1 in disease states. Our study has several strengths. This study included a large population of women with ovulatory cycles for changes in VEGF in serum, plasma, and urine. The inclusion of sFLT-1 was novel to more fully characterize important markers of angiogenesis during the menstrual cycle. Further, this study evaluated associations between VEGF and sFLT-1 with hormonal changes using models that account for the unique variation of reproductive hormones across the menstrual cycle. In addition, we were able to adjust for variation in individual menstrual cycles. Specifically, we confirmed detection of the LH surge with the aid of fertility monitors, allowing for proper scheduling of biospecimen collection, decreasing the likelihood of mis-categorizing the cycle in which a measurement was collected. As participants contributed multiple consecutive samples for analysis of plasma, serum, and urine throughout the menstrual cycle, we were able to thoroughly capture changing hormonal profiles that may have been related to VEGF and sFLT-1 concentrations. Our strict exclusion criteria also ensured that our results were not impacted by recent pregnancy, infection, or use of hormonal medications. The generalizability of our findings may be limited by the inclusion of women only from one geographic location and generally non-obese women. In addition, the cohort consisted of mostly White women, further limiting the generalizability of the study. Also, accordingly with the aim of this investigation, these data cannot address potential variation attributable to disease states (e.g. anovulation, endometriosis, recurrent pregnancy loss, and more.) Given that we found no changes in VEGF or sFLT-1 measured in serum, plasma, or urine across the menstrual cycle in healthy, ovulatory women, we conclude that circulating VEGF and sFLT-1 are not useful peripheral biomarkers for endometrial remodeling across healthy menstrual cycles. It remains a need to identify a noninvasive marker of endometrial function that may contribute to our understanding of fertility, pregnancy, and their complications.

The endometrium is in a constant state of transformation, undergoing repair, regeneration, and regression at critical time points during the menstrual cycle in preparation for pregnancy. These changes in the endometrium are directed by steroid hormones and coordinated with factors that promote angiogenesis 1 . Reproductive and gynecologic complications such as abnormal uterine bleeding 2 – 5 , failed implantation 6 , recurrent pregnancy loss 7 , preeclampsia 8 , and abnormal placentation 9 – 11 may result from dysregulation of endometrial transformation, and consequently placentation in the case of pregnancy, but little is known about their cause and prevention. Accordingly, a simple, noninvasive marker of endometrial dysfunction suitable for large-scale studies of such reproductive complications is needed to assess endometrial function and may provide a biologic mechanism for common reproductive and gynecologic complications and obstetric complications that may lead to maternal and neonatal morbidity. However, it is unknown if circulating vascular endothelial growth factor (VEGF) and soluble fms-like tyrosine kinase-1 (sFLT-1) change throughout the menstrual cycle and how they may contribute to or signal pathologic conditions. Presently, evaluation of the uterus is limited to invasive biopsy, various imaging modalities, or surgical procedures. Investigation of systemic angiogenic markers during the menstrual cycle may lead to better understanding of disease processes and avoid the need to undergo invasive and expensive procedures. . VEGF signaling is essential for endometrial proliferation and differentiation 12 and regulates trophoblast invasion during implantation 13 . Numerous studies relying on tissue obtained from invasive procedures indicate that ovarian steroids may regulate VEGF synthesis, resulting in changes in VEGF across the cycle 14 – 16 . At the tissue level, VEGF is upregulated during menstruation, in the early proliferative phase, and during the secretory phase, facilitating the development of spiral arteries, suggesting VEGF as a biomarker of endometrial remodeling. For large-scale studies which are necessary to examine rare complications on a population level, however, characterizing changes in peripheral VEGF, and its circulating antagonist receptor, sFLT-1, across the normal menstrual cycle is an important step to investigate its use as a useful biomarker of endometrial health. However, current evidence is conflicting regarding cyclic changes in circulating VEGF 17 – 21 . In addition, the optimal biospecimen for measurement of VEGF is unclear, as studies have used blood serum and plasma, as well as urine. Since VEGF is released from platelets and leukocytes during blood collection and clotting, higher VEGF is detected in serum 22 , indicating a unique need to characterize VEGF in each of these difference specimen types when investigating potential relationships with reproductive hormones across the menstrual cycle. Therefore, the aim of our study was to explore the relationship of cyclical changes in VEGF and its receptor, sFLT-1, measured in plasma, serum, and urine with serum estradiol, progesterone, follicle stimulating hormone (FSH), and luteinizing hormone (LH) across the menstrual cycle in a large, well-characterized sample of healthy, regularly menstruating women. As studies at the tissue level have indicated that hormonal changes may regulate angiogenic factors, we hypothesize that circulating VEGF and sFLT-1 will fluctuate throughout the menstrual cycle. These data will inform the utility of circulating VEGF and its receptor in large-scale studies of reproductive complications potentially arising from endometrial dysfunction.