Relationships between reproductive hormones and maternal pregnancy physiology in women conceiving with or without in vitro fertilization

Abstract

We evaluated maternal pregnancy adaptations and their relationships with circulating hormones in women who conceived with or without in vitro fertilization (IVF). Pregnancies were grouped by corpus luteum (CL) number: 1 CL with physiological plasma relaxin concentration (P_RLN; spontaneous pregnancies); 0 CL without circulating RLN (programmed cycles); >1 CL with elevated P_RLN (ovarian stimulation). Major findings were that declines in plasma osmolality (P_osm) and plasma sodium concentration (${\mathrm{P}}_{\mathrm{N}{\mathrm{a}}^{\mathrm{+}}}$) were comparable in the 1 CL and 0 CL cohorts, correlated with plasma estradiol and progesterone concentrations but not P_RLN; gestational declines in plasma uric acid (UA) concentration (P_UA) were attenuated after IVF, especially programmed cycles, partly because of subdued increases of renal UA clearance; and P_RLN and cardiac output (CO) were inversely correlated when plasma estradiol concentration was below ∼2.5 ng/mL but positively correlated above ∼2.5 ng/mL. Unexpectedly, P_RLN and plasma sFLT1 (P_sFLT1) were directly correlated. Although P_sFLT1 and CO were not significantly associated, CO was positively correlated with plasma placental growth factor (PLGF) concentration after the first trimester, particularly in women who conceived with 0 CL. Major conclusions are that 1) circulating RLN was unnecessary for gestational falls in P_osm and ${\mathrm{P}}_{\mathrm{N}{\mathrm{a}}^{\mathrm{+}}}$; 2) P_RLN and CO were inversely correlated during early gestation, suggesting that P_RLN in the lower range may have contributed to systemic vasodilation, whereas at higher P_RLN RLN influence became self-limiting; 3) evidence for cooperativity between RLN and estradiol on gestational changes in CO was observed; and 4) after the first trimester in women who conceived without a CL, plasma PLGF concentration was associated with recovery of CO, which was impaired during the first trimester in this cohort.

INTRODUCTION

In vitro fertilization (IVF) has been associated with increased occurrence of pathological pregnancy outcomes. In particular, frozen embryo transfer (FET) was linked to higher rates of preeclampsia (PE), thus implicating the freezing procedure (reviewed in Ref. 1). More recently, however, the absence of a corpus luteum (CL) was specifically implicated in the elevated risk of PE and PE with severe features among women who conceived by autologous FET in programmed cycles, which precluded the formation of a CL (1, 2). This increased PE risk was not observed in women conceiving by autologous FET in a (modified) natural or mildly stimulated cycle or in controlled ovarian stimulation cycles with fresh embryo transfer in which one or more CLs develop. PE risk following these IVF protocols was similar to spontaneous conceptions (1–3).

Other than a few classic hormones including estradiol, progesterone, and relaxin, the complete repertoire of hormones and other factors secreted by the CL during pregnancy is largely unknown. In the gravid rat model, relaxin plays a key role in the gestational changes of renal and cardiovascular function as well as plasma osmolality, since they were abrogated by neutralization or elimination of circulating relaxin by antibodies or ovariectomy, respectively, during early to midterm pregnancy (4, 5). To test a role for CL factors in human pregnancy, we evaluated women who conceived with IVF (6). This “experimental model” afforded the opportunity to investigate pregnancy physiology in the absence of a CL (programmed cycles) or in the presence of multiple CLs (controlled ovarian stimulation cycles) compared with a control group of women conceiving without IVF and with one CL (spontaneous singleton pregnancies) (6). We observed significant attenuation of the gestational changes in cardiovascular function, particularly in the first trimester of women who conceived in the absence of a CL, and of circulating CL factors like relaxin (2, 7). As mentioned above, these women also demonstrated higher rates of PE and PE with severe features (2). In contrast, the women who conceived by controlled ovarian stimulation were comparable to spontaneous conceptions with regard to PE risk and cardiovascular function during early pregnancy (2, 7).

The goal of the present work was twofold. First, we set out to explore additional physiological variables that we measured in the women who conceived with and without IVF, including plasma osmolality, sodium, and uric acid and renal uric acid clearance, in addition to the presence or absence of microalbuminuria as a sensitive biomarker of vascular dysfunction. Previous work by us and others suggested a role for the CL in the osmoregulatory changes of early human pregnancy (8–10) but not in mid- to late pregnancy (8). Second, we wanted to investigate associations between circulating concentrations of reproductive and other hormones with plasma osmolality and sodium, as well as with the cardiovascular variables cardiac output (CO) and systemic vascular resistance (SVR) (7, 11). Many of the findings that emerged from this work were counterintuitive and unanticipated, underscoring the fundamentally inscrutable and illusive complexity of human pregnancy.

METHODS

Participants

After written informed consent, participants were enrolled in this study, which was approved by the University of Florida Institutional Review Board. Women intending to conceive through IVF were identified by reproductive endocrinology and infertility specialists. Women planning to become pregnant without assisted reproduction were recruited through advertisement. Thus, the three participant groups were composed of women conceiving 1) spontaneously without IVF (singleton pregnancy with 1 CL); 2) by IVF and transfer of frozen embryo(s) produced with autologous or donor eggs or fresh embryo(s) derived from donor eggs (absent CL); or 3) by IVF and fresh embryo transfer following controlled ovarian stimulation (multiple CL).

In brief, baseline measurements of plasma osmolality, sodium, and uric acid were made before pregnancy in the absence of circulating CL factors (follicular phase or last 4 days of leuprolide suppression). Then, participants were studied six times during pregnancy: 5.8 ± 0.1, 8.3 ± 0.1, 11.6 ± 0.1, 15.2 ± 0.1, 24.0 ± 0.1, and 33.6 ± 0.1 wk of gestation, as well as 45.9 ± 1.8 wk after delivery (means ± SE). The number of women investigated at each time point is presented in Supplemental Table S4, which also provides means ± SE of all measured variables (all Supplemental Material is available at https://doi.org/10.6084/m9.figshare.15127500). Women included in this analysis had at least three pregnancy visits, in addition to the prepregnancy baseline. After fasting overnight and abstinence from pain medications, caffeinated products, and alcohol during the preceding 24 h, participants arrived at the Clinical Research Center at ∼8:00 AM, where they underwent physiological evaluations over a period of ∼5 h. Fasting blood samples were obtained between 9:00 and 9:30 AM, and, except for the occasional difficult phlebotomy, blood for the hormones and other factors reported here was obtained immediately within the first minute or two of needle insertion. In addition, 24-h urines were collected at each visit.

Assisted Reproduction Protocols

Standard protocols were employed for fresh IVF cycles. Identical programmed cycle protocols were implemented for donor oocyte recipient fresh embryo transfer and FET and autologous oocyte FET. In fresh autologous IVF cycles using controlled ovarian stimulation, CL number was estimated by the number of retrieved eggs. In programmed cycles, absent CL was verified by ultrasound and by undetectable concentration of plasma relaxin throughout pregnancy (11).

Regardless of protocol, patients were instructed to avoid caffeine, alcohol, herbal supplements, and over-the-counter medications and to take a multivitamin containing at least 0.4 mg of folic acid. All three protocols were begun with combined oral estrogen-progesterone contraceptives (OCPs) starting on day 3 of the menstrual cycle taken for 14–40 days (on average 21 days) depending on when the menstrual cycle coincided with IVF cycle dates.

For the microdose flare protocol, patients started 0.5 mg of dexamethasone orally and 81 mg of aspirin (ASA) orally daily 1 wk before the start of luprolide. OCPs were discontinued after 14–21 days. Four days later, 40 µg of subcutaneous luprolide twice daily was started. Recombinant (r)FSH (450 IU sc daily) was started the following day for 5–12 days pending ovarian response on serial ultrasounds and estradiol levels. When appropriate response was achieved, rFSH, luprolide, and dexamethasone were stopped, and 10,000 IU of human chorionic gonadotropin (hCG) was given subcutaneously to trigger final oocyte maturation.

The gonadotropin-releasing hormone (GnRH) agonist suppression protocol was started with 1 mg of subcutaneous liprolide daily beginning after pretreatment with OCPs. The same doses as described above of dexamethasone and ASA were started simultaneously with luprolide. OCPs were stopped 3 days after starting luprolide. Once adequate suppression was achieved, usually after 10–14 days, 75–375 IU of rFSH subcutaneously and 75 IU of FSH/luteinizing hormone (LH) subcutaneously were begun and luprolide was decreased to 0.25 mg subcutaneously daily. Doses of gonadotropins were adjusted as necessary based on follicular response determined by ultrasound and estradiol levels. After 8–12 days of stimulation, 5,000–10,000 IU of hCG was given subcutaneously to trigger final oocyte maturation. FSH, dexamethasone, and luprolide were stopped at this time.

For the antagonist protocol, 10 days before stimulation, the above-mentioned doses of dexamethasone and ASA were started. rFSH (100–450 IU sc) was given for 5–12 days, and 75 IU of subcutaneous FSH/LH was added if LH was ≤ 1.0 IU/ml on the day of the suppression check. When follicles reached ∼14 mm in size, the GnRH antagonist antigon and an additional 75 IU of subcutaneous FSH/LH were started daily. After adequate stimulation, 10,000 IU of subcutaneous hCG or a 2-mg subcutaneous luprolide trigger was given to stimulate final oocyte maturation.

Egg retrievals were performed in all cases ∼35 h posttrigger. All three protocols followed the same steps for oocyte retrieval. On the day of retrieval, Medrol (4 mg 4 times/day orally) and doxycycline (100 mg twice a day orally) were started and continued through the day of embryo transfer. On the day after retrieval, intramuscular progesterone 1 mL/50 mg was started nightly. Both progesterone and ASA were continued until 8 wk of gestation. All patients were given 10 mg of Valium orally 1 h before embryo transfer.

The frozen embryo transfer protocol started with OCPs on day 3 of the menstrual cycle for 14–40 days. Three days before discontinuation of OCPs, 81 mg of oral ASA and 1 mg of subcutaneous luprolide were started. Once adequate suppression was achieved, a 0.1-mg estradiol patch was placed and changed every other day. Luprolide was decreased to 0.25 mg subcutaneously daily at this time. On days 7–8 of estrogen therapy, dosing was increased to two patches every other day, on days 9–12 dosing was increased to three patches every other day, and on days 13–14 it was increased to four patches every other day. Pending adequate response to estrogen, determined by blood estrogen levels and ultrasound, luprolide was stopped and micronized progesterone 200 mg vaginally twice a day and intramuscular progesterone 1 mL/50 mg nightly were started. The day before embryo transfer, Medrol (4 mg 4 times/day orally) was started and doxycycline (100 mg twice a day) for 6 days. Estradiol patches, ASA, and progesterone were continued until 10 weeks of gestation. The protocol for autologous frozen embryo transfer was identical to the egg donor embryo transfer protocol regardless of fresh or frozen donor embryo.

Assays

Plasma osmolality was measured by the vapor pressure method (Wescor Inc., VAPRO 5520) and plasma sodium by the Piccolo Xpress Chemistry Analyzer (Abbott Point of Care, Princeton, NJ). Plasma and urine uric acid were determined through direct measurement with Kit No. DIAU-250 (Bioassay Systems, Hayward, CA). Urinary microalbumin-to-creatinine ratio was measured in the University of Florida Medical Laboratory-Shands Hospital with a Beckman Coulter AU5800.

We previously reported the circulating concentrations of reproductive and other hormones, cardiovascular function, and their methods of measurement in our participant cohorts conceiving with and without IVF (7, 11).

Statistical Analyses

Numerical characteristics are described by means ± SE [range]. Nonparametric ANOVA (Kruskal–Wallis) test was used to compare numerical variables between two or more groups. Post hoc stepdown analysis with nonparametric Wilcoxon rank sum test was used with Bonferroni correction. Categorical characteristics were described with frequencies (proportions) and compared between groups with the Fisher’s exact test. Pearson correlation coefficient (r) was used to assess linear relationship between numerical characteristics. Multiple linear regression was applied to the effect of changes in plasma concentrations of estradiol and relaxin from prepregnant baseline, as well as their interaction, on the change in cardiac output from prepregnant baseline during the first trimester in the spontaneous pregnancy cohort. Linear mixed model with cubic splines for gestation weeks was fitted to evaluate trends and compare changes over gestation. The full model included group, gestational week, and group × gestational week interactions. Linear mixed models account for the correlation between repeated measurements made in the same women at multiple time points and include random effects to model the correlation and allow for missing observations to occur during the study. Significance level was set at the 0.05 level. SAS (9.4) and R (3.6.1) were used for the analysis.

RESULTS

Participant Characteristics, Infertility Diagnoses, Obstetrical and Neonatal Outcomes

There were 19–24 participants per group. Participant race and ethnicity, parity, and history of smoking and hypertensive disease of pregnancy were similar among the three groups. Maternal age and body mass index (BMI) were 4 yr greater and 3–4 kg/m² higher, respectively, in the 0 CL compared with the single and multiple CL cohorts (both P < 0.05; Table 1). Diminished ovarian reserve and male factor infertility, respectively, were more frequent in the 0 CL and multiple CL cohorts (Supplemental Table S1). A greater number of pathological obstetrical outcomes were observed in women who conceived by IVF; however, the study was underpowered for detecting statistical differences in pregnancy outcomes (Supplemental Table S2). Not unexpectedly, the number of twin pregnancies was significantly greater and gestational age at delivery and newborn weight significantly lower in IVF compared with spontaneously conceived pregnancies (Supplemental Table S3).

Table 1.

Participant demographic and clinical characteristics

	0 CL	1 CL	Multiple CL	P Value
N	24	22	19
Maternal age (at 5–6 gestational wk), yr	36 ± 1 [29–46]	32 ± 1 [23–45]	32 ± 1 [26–41]	0.03
BMI (prepregnancy), kg/m2	27 ± 1 [18–36]	24 ± 1 [18–43]	23 ± 1 [18–38]	0.02
Body weight (prepregnancy), kg	74 ± 3 [47–101]	66 ± 3 [44–115]	65 ± 3 [48–100]	0.16
Height, cm	165 ± 2 [154–183]	165 ± 1 [152–172]	167 ± 2 [157–182]	0.84
Participant race, n (%)
White	20 (83)	20 (91)	16 (84)	0.54
American Indian	0	0	0
Alaska Native	0	0	0
Asian	2 (8)	1 (5)	0
Black or African American	2 (8)	1 (5)	1 (5)
Other			2 (10)
Nulliparity, n (%)	11(46)	11 (50)	12 (63)	0.55
Maternal smoking, n (%)	2 (10)	1 (4)	0	0.77
Hypertensive disease in prior pregnancy, n	0	0	0	NA

Values are means ± SE [ranges] or n (%) for N participants. BMI, body mass index; CL, corpus luteum; NA, not applicable. Boldface indicates significance.

Plasma Osmolality and Sodium Concentration in Pregnancies Conceived Spontaneously or by IVF: Lack of Association with Plasma Relaxin or hCG Concentrations

Unexpectedly, the declines in plasma osmolality during the first trimester in women conceiving with and without IVF were comparable (Fig. 1, Supplemental Table S4). The full mixed model analysis of plasma osmolality during the first trimester yielded a nonsignificant group × time interaction (P = 0.888; Fig. 1A). At 10–12 wk gestation, nonparametric ANOVA revealed a nonsignificant difference in the change of plasma osmolality from prepregnancy baseline between >1 CL and 1 CL + 0 CL (P = 0.111; Fig. 1B), and the areas under the curves over the first trimester for >1 CL versus 1 CL + 0 CL were also not significantly different (P = 0.665; Fig. 1B). As shown in Tables 2 and 3, neither plasma relaxin nor log relaxin concentration was significantly correlated with plasma osmolality during the first trimester in women who conceived spontaneously (1 CL) or by fresh embryo transfer (>1 CL). A similar lack of associations was observed when the two participant groups were analyzed together (relaxin vs. osmolality: r = 0.0161, r² = 0.000, P = 0.863; log relaxin vs. osmolality: r = 0.003, r² = 0.000, P = 0.973). In addition, no significant correlations were observed between plasma relaxin or log relaxin concentration and the change in plasma osmolality from prepregnant baseline for the 1 CL and >1 CL participant cohorts combined during the first trimester (data not shown). The inverse relationships between plasma hCG or log hCG concentration and plasma osmolality observed in the first trimester were also not significant (Tables 2 and 3).

Figure 1.

Absolute mean (A) and delta (Δ) mean (B) plasma osmolality concentrations before, during, and after pregnancy in women who conceived spontaneously [1 corpus luteum (CL)] or via in vitro fertilization (IVF) (0 CL and >1 CL). In A, over the first trimester, the full mixed model analysis yielded group P = 0.549, time P < 0.001, and group × time P = 0.888. In B, nonparametric ANOVA comparing >1 CL vs. 1 CL + 0 CL at 10–12 wk of gestation, P = 0.111, and area under the curves over the first trimester for >1 CL vs. 1 CL + 0 CL, P = 0.665. BP, before pregnancy in the follicular phase or last 4 days of leuprolide suppression; PP, postpartum. Means ± SE; n, no. subjects.

Table 2.

Correlations between plasma hormone concentrations and plasma osmolality

Comparison/Group	N	r	r 2	P Value
RLN vs. Osm
0 CL
1 CL	63	0.020	0.000	0.878
>1 CL	54	−0.002	0.000	0.990
hCG vs. Osm
0 CL	68	−0.070	0.005	0.568
1 CL	63	−0.158	0.025	0.217
>1 CL	54	−0.27	0.073	0.048
E2 vs. Osm
0 CL	90	−0.254	0.064	0.016
1 CL	85	−0.367	0.135	0.001
>1 CL	73	−0.460	0.212	<0.001
P4 vs. Osm
0 CL	90	−0.397	0.157	<0.001
1 CL	85	−0.465	0.216	<0.001
>1 CL	73	−0.427	0.182	<0.001
PRA vs. Osm
0 CL	92	−0.281	0.079	0.007
1 CL	85	−0.335	0.112	0.002
>1 CL	69	−0.326	0.106	0.006
Aldo vs. Osm
0 CL	91	−0.313	0.098	0.003
1 CL	84	−0.377	0.142	<0.001
>1 CL	71	−0.381	0.145	0.001

Correlations included the measurements made before pregnancy and during the first trimester except for relaxin (RLN) and human chorionic gonadotropin (hCG), which included only the first trimester because they were undetectable before pregnancy. Aldo, aldosterone; E₂, estradiol; Osm, plasma osmolality; PRA, plasma renin activity; P4, progesterone. Boldface indicates significance. N, No. samples analyzed.

Table 3.

Correlations between log plasma hormone concentrations and plasma osmolality

Comparison/Group	N	r	r 2	P Value
Log RLN vs. Osm
0 CL
1 CL	63	0.046	0.002	0.721
>1 CL	54	−0.035	0.001	0.802
Log hCG vs. Osm
0 CL	68	−0.053	0.003	0.668
1 CL	63	−0.161	0.026	0.206
>1 CL	54	−0.259	0.067	0.058
Log E2 vs. Osm
0 CL	90	−0.483	0.233	<0.001
1 CL	85	−0.545	0.297	<0.001
>1 CL	73	−0.525	0.276	<0.001
Log P4 vs. Osm
0 CL	90	−0.487	0.238	<0.001
1 CL	85	−0.530	0.281	<0.001
>1 CL	73	−0.566	0.320	<0.001
Log PRA vs. Osm
0 CL	90	−0.343	0.118	0.001
1 CL	84	−0.390	0.152	<0.001
>1 CL	68	−0.418	0.175	<0.001
Log Aldo vs. Osm
0 CL	91	−0.388	0.151	<0.001
1 CL	84	−0.445	0.198	<0.001
>1 CL	71	−0.615	0.378	<0.001

Correlations included the measurements made before pregnancy and during the first trimester except for RLN and hCG, which included only the first trimester because they were undetectable before pregnancy. Aldo, aldosterone; E₂, estradiol; hCG, human chorionic gonadotropin; Osm, plasma osmolality; PRA, plasma renin activity; P4, progesterone; RLN, relaxin. Boldface indicates significance. N, No. samples analyzed.

As expected based on the findings for plasma osmolality, comparable declines in plasma sodium concentration were observed during the first trimester in women conceiving with and without IVF (Fig. 2, Supplemental Table S4). In Fig. 2A, the group × time interaction was not significant for plasma sodium during the first trimester (P = 0.783). In contrast, at 10–12 wk of gestation, nonparametric ANOVA revealed a borderline significant difference in the change of plasma sodium from prepregnancy baseline between >1 CL and 1 CL + 0 CL (P = 0.058; Fig. 2B), and the areas under the curves for >1 CL versus 1 CL + 0 CL were significantly different (P = 0.024; Fig. 2B). Nevertheless, the relationships between plasma relaxin or log relaxin and sodium concentration were not significant (Supplemental Tables S5 and S6). When the 1 CL and >1 CL groups were analyzed together, the correlations were weak at best (relaxin vs. sodium: r = 0.210, r² = 0.044, P = 0.026; log relaxin vs. sodium: r = 0.057, r² = 0.003, P = 0.548. No significant correlations were observed between plasma hCG or log hCG and sodium concentration, either (Supplemental Tables S5 and S6).

Figure 2.

Absolute mean (A) and delta (Δ) mean (B) plasma sodium concentrations before, during, and after pregnancy in women who conceived spontaneously [1 corpus luteum (CL)] or via in vitro fertilization (IVF) (0 CL and >1 CL). In A, over the first trimester, the full mixed model analysis yielded group P = 0.056, time P < 0.001, group × time P = 0.783. In B, nonparametric ANOVA comparing >1 CL vs. 1 CL + 0 CL at 10–12 wk of gestation, P = 0.058, and area under the curves for >1 CL vs. 1 CL + 0 CL, P = 0.024. BP, before pregnancy in the follicular phase or last 4 days of leuprolide suppression; PP, postpartum. Means ± SE; n, No. subjects.

Plasma Osmolality and Sodium Concentration in Pregnancies Conceived Spontaneously or by IVF: Association with Plasma Progesterone and Estradiol Concentrations

There were modest, but significant, inverse correlations between plasma estradiol or log estradiol concentration and osmolality for each cohort (Tables 2 and 3) and for the three cohorts combined during the first trimester (r² = 0.099 and 0.256, respectively, both P < 0.001; Fig. 3A). Similarly, there were modest, but significant, inverse correlations between plasma progesterone or log progesterone concentration and osmolality for each participant cohort (Tables 2 and 3) and for the three cohorts combined in the first trimester (r² = 0.104 and 0.259, respectively, both P < 0.001; Fig. 3B). Again, however, the correlations between plasma estradiol or progesterone (or the logs thereof) and sodium concentration for each cohort were generally weaker but mostly statistically significant (Supplemental Tables S5 and S6). As expected, the direct correlation between plasma progesterone and estradiol concentrations was high for the three cohorts combined during the first trimester (log estradiol vs. log progesterone: r² = 0.764, P < 0.001). Unexpectedly, the direct relationship between plasma osmolality and sodium concentration was weak (r² = 0.108, P < 0.001).

Figure 3.

Correlations of plasma estradiol and progesterone concentrations with plasma osmolality before pregnancy (follicular phase) and during the first trimester in women conceiving spontaneously or by in vitro fertilization (IVF). A: plasma estradiol (log pg/mL) concentration vs. plasma osmolality (mosmol/kgH₂O): plasma osmolality = 287.9 − 1.77 × log estradiol; r = −0.506, r² = 0.256, P < 0.001. B: plasma progesterone (log ng/mL) vs. plasma osmolality (mosmol/kgH₂O): plasma osmolality = 279.9 − 1.40 × log progesterone; r = −0.509, r² = 0.259, P < 0.001. Programmed IVF cycles [without corpus luteum (CL)], controlled ovarian stimulation cycles (multiple CLs), and spontaneously conceived pregnancies (1 CL).

Plasma Osmolality and Sodium Concentration in Pregnancies Conceived Spontaneously or by IVF: Association with Plasma Renin Activity and Aldosterone

Plasma osmolality correlated inversely and significantly with plasma renin activity (PRA) and log PRA and with plasma aldosterone and log aldosterone concentration for each of the three participant groups during the first trimester (Tables 2 and 3). The relationships between PRA or aldosterone and plasma sodium concentration were considerably weaker, and the majority were nonsignificant (Supplemental Table S5). Weaker associations were also observed for log PRA or log aldosterone and plasma sodium concentration, although in this case most were significant (Supplemental Table S6).

Plasma Uric Acid and Renal Uric Acid Clearance

During normal pregnancy, renal clearance of uric acid rises because of gestational increases of glomerular filtration rate, and hence filtered load, decreases in proximal tubular reabsorption, or both (see discussion for details). To our knowledge, these variables have not been reported for women who conceived by IVF. The pattern of change in plasma uric acid concentration during pregnancy was notably different for the women who conceived by IVF using programmed cycles (0 CL) compared with those who conceived by controlled ovarian stimulation (>1 CL) or spontaneously (1 CL) (Fig. 4A, Supplemental Table S4). There was a significant group × time interaction between the 0 CL and the 1 CL and >1 CL cohorts combined throughout pregnancy (P = 0.028), the latter two cohorts being comparable. In addition, there were significant differences between individual time points in the first trimester relative to prepregnancy values for the 1 and >1 CL cohorts (P < 0.0167) but not for the 0 CL cohort. Thus, the decline of plasma uric acid in the 0 CL cohort was attenuated during the first trimester.

Figure 4.

Absolute mean (A) and delta (Δ) mean (B) plasma uric acid concentrations before, during, and after pregnancy in women who conceived spontaneously [1 corpus luteum (CL)] or via vitro fertilization (IVF) (0 CL and >1 CL). A: *P < 0.0167 vs. before pregnancy in the follicular phase or last 4 days of leuprolide suppression (BP). B: **P < 0.05 comparing all 3 cohorts (nonparametric ANOVA); P ≤ 0.001 comparing area under the curves of all 3 cohorts either during the first trimester or throughout pregnancy. PP, postpartum. Means ± SE; n, No. subjects.

However, closer inspection of the change in plasma uric acid suggested that, relative to prepregnancy baseline, the decline in plasma uric acid for the >1 CL cohort was also subdued during the first trimester compared with the 1 CL group (Fig. 4B). In addition, plasma uric acid concentration during pregnancy within each participant cohort was similar whether adverse pregnancy outcomes were included in the analysis or not, with one exception: plasma uric acid concentration reached a lower peak at 32–35 wk of gestation particularly in the 0 CL group after the adverse pregnancy outcomes were excluded from the analysis: 5.9 ± 0.3 mg/dL (vs. 6.2 ± 0.3 mg/dL for the normal and pathological pregnancy outcomes combined). Nevertheless, in the 0 CL group at 32–35 wk of gestation, the plasma uric acid concentration exceeded the 1 CL cohort by ±1 to ±2 standard deviations, regardless of the pregnancy outcome.

One possible explanation for the subdued decrease of plasma uric acid concentration in the two IVF cohorts, especially for the 0 CL group, was an attenuated rise in renal uric acid clearance (Fig. 5, Supplemental Table S4). Although there was not a significant group × time interaction in the first trimester among the three participant cohorts, only the women with spontaneous pregnancies showed a significant increase in uric acid clearance in the first trimester relative to prepregnancy baseline (Fig. 5A; P < 0.0167 for gestational weeks 7–9 and 10–12). Moreover, the area under the curve for the change in uric acid clearance from prepregnancy baseline was significantly greater for 1 CL versus 0 and >1 CL combined during the first trimester (P < 0.05) and throughout pregnancy (P < 0.01), the area under the curve for the two IVF groups being comparable (Fig. 5B). There was a modest, albeit significant, inverse relationship between the renal clearance and plasma concentration of uric acid in the prepregnancy period and throughout pregnancy for the three cohorts combined (r² = 0.127, P < 0.001). Interestingly, this relationship was stronger for the 0 CL and 1 CL cohorts when analyzed separately (r² = 0.151 and 0.227, respectively, both P < 0.001) and much weaker for the >1 CL cohort (r² = 0.021, P = not significant).

Figure 5.

Absolute mean (A) and delta (Δ) mean (B) renal uric acid clearance before, during, and after pregnancy in women who conceived spontaneously [1 corpus luteum (CL)] or via vitro fertilization (IVF) (0 CL and >1 CL). A: *P < 0.0167 vs. before pregnancy in the follicular phase or last 4 days of leuprolide suppression (BP). B: †P < 0.05 comparing 1 CL vs. 0 CL and >1 CL combined (nonparametric ANOVA); P < 0.05 and < 0.01 comparing area under the curves of 1 CL vs. 0 CL and >1 CL combined during the first trimester and throughout pregnancy, respectively. PP, postpartum. Means ± SE; n, No. subjects.

Microalbuminuria

Microalbuminuria was determined at three time points: 1) prepregnancy, as the baseline control; 2) early pregnancy, when gestational renal hyperfiltration typically transpires, which could exacerbate any underlying albuminuria; and 3) postpartum, in order to assess whether either IVF might negatively affect kidney function remote from pregnancy. Microalbuminuria was detected in 5/17, 7/13, and 3/15 women who conceived by programmed cycles, controlled ovarian stimulation cycles, or spontaneously, respectively (Supplemental Table S7). Thus, 12/30 or 40% of participants who conceived by assisted reproduction and 3/15 or 20% of participants who conceived spontaneously manifested albuminuria. The degree of microalbuminuria was within the normal range in 50% of cases (0–29 mg/g creatinine). In considering the potential effect of IVF pregnancy on the development of microalbuminuria postpartum, 5 of 30 (17%) and 2 of 15 (13%) women, respectively, who conceived by IVF or spontaneously had detectable albuminuria after pregnancy.

Hormones, Angiogenic Growth Factors, Cardiac Output, and Systemic Vascular Resistance

Relaxin and cardiac output.

Unexpectedly, there were inverse relationships between circulating relaxin and cardiac output (CO) especially at 5–6 gestational wk among the women who conceived spontaneously (1 CL; Table 4, Fig. 6A). These correlations were relaxin vs. CO: r = −0.600, r² = 0.360, P = 0.004; log relaxin vs. CO: r = −0.609, r² = 0.371, P = 0.003; Δrelaxin vs. ΔCO prepregnant to 5–6 wk: r = −0.637, r² = 0.406, P = 0.002; Δlog relaxin vs. ΔCO prepregnant to 5–6 wk: r = −0.615, r² = 0.378, P = 0.003. Thus, in the spontaneous conception cohort, on average ∼38% of the variability in CO could be explained by the plasma relaxin concentration. For the women who conceived by controlled ovarian stimulation (>1 CL), the only relationship that approached significance was observed at 10–12 gestational wk for relaxin versus CO (r = −0.412, r² = 0.169, P = 0.090). It should be pointed out that plasma relaxin concentration in this cohort ranged from values observed in the spontaneous conception group up to approximately six times higher (11).

Table 4.

Correlation between plasma relaxin concentration and cardiac output

Comparison	Group	N	R	r 2	P Value
Spontaneous conceptions (1 CL): 5–6 gestational wk
RLN vs. CO	1 CL	21	−0.600	0.360	0.004
Log RLN vs. CO	1 CL	21	−0.609	0.371	0.003
RLN vs. ΔCO	1 CL	21	−0.637	0.406	0.002
Log RLN vs. ΔCO	1 CL	21	−0.615	0.378	0.003
Spontaneous conceptions (1 CL): 7–9 gestational wk
RLN vs. CO	1 CL	21	−0.383	0.147	0.087
Log RLN vs. CO	1 CL	21	−0.434	0.189	0.049
RLN vs. ΔCO	1 CL	21	−0.301	0.091	0.184
Log RLN vs. ΔCO	1 CL	21	−0.346	0.120	0.124
Spontaneous conceptions (1 CL): 10–12 gestational wk
RLN vs. CO	1 CL	21	−0.422	0.178	0.057
Log RLN vs. CO	1 CL	21	−0.411	0.169	0.064
RLN vs. ΔCO	1 CL	21	−0.386	0.149	0.084
Log RLN vs. ΔCO	1 CL	21	−0.385	0.148	0.085
Spontaneous conceptions (1 CL): 14–16 gestational wk
RLN vs. CO	1 CL	21	−0.218	0.048	0.342
Log RLN vs. CO	1 CL	21	−0.037	0.001	0.873
RLN vs. ΔCO	1 CL	21	−0.160	0.026	0.488
Log RLN vs. ΔCO	1 CL	21	−0.020	0.000	0.933
Spontaneous conceptions (1 CL): 23–25 gestational wk
RLN vs. CO	1 CL	20	−0.372	0.139	0.106
Log RLN vs. CO	1 CL	20	−0.279	0.078	0.234
RLN vs. ΔCO	1 CL	20	−0.499	0.249	0.025
Log RLN vs. ΔCO	1 CL	20	−0.479	0.229	0.033
Spontaneous conceptions (1 CL): 32–35 gestational wk
RLN vs. CO	1 CL	19	−0.451	0.203	0.053
Log RLN vs. CO	1 CL	19	−0.415	0.172	0.077
RLN vs. ΔCO	1 CL	19	−0.626	0.392	0.004
Log RLN vs. ΔCO	1 CL	19	−0.576	0.332	0.010

Because prepregnant plasma relaxin (RLN) concentration was undetectable (<7.8 pg/mL), the plasma RLN concentration measured at 5–6 gestational wk was used. CL, corpus luteum; CO, cardiac output. Boldface indicates significance. N, No. subjects.

Figure 6.

Correlations of the change in cardiac output (CO), systemic vascular resistance (SVR), and sFLT1 from prepregnant baseline with plasma relaxin concentration at 5−6 gestational wk in women who conceived spontaneously. A and B: plasma relaxin concentration (pg/mL) vs. CO (ΔmL/min; A): ΔCO = 2270.77 − 2.39 × Δrelaxin; r = −0.637, r² = 0.406, P = 0.002 and SVR (ΔmmHg/mL·min⁻¹; B): ΔSVR = 7.441 + 0.008 × Δrelaxin; r = 0.627, r² = 0.393, P = 0.002. C: log plasma relaxin vs. ΔsFLT1 concentration (Δpg/mL): ΔsFLT1= −1226.7 + 234.5 × log relaxin; r = 0.452, r² = 0.204, P = 0.040. Because prepregnant plasma relaxin concentration was undetectable (<7.8 pg/mL), the plasma relaxin concentration measured at 5–6 gestational wk is plotted on the abscissa.

Relaxin, estradiol, and cardiac output.

Multiple linear regression analysis suggested a cooperative interaction between plasma estradiol and relaxin on the rise of cardiac output during the first trimester in the spontaneously conceiving cohort (Fig. 7). In particular, the paradoxical, negative relationship between plasma relaxin concentration and cardiac output noted above changed to a positive association when plasma estradiol reached, on average, ∼2.5 ng/mL, which occurred during the first trimester in this participant group (11).

Figure 7.

Multiple linear regression model provides evidence for cooperation between plasma relaxin and estradiol on cardiac output during the first trimester. At a plasma estradiol concentration in the range of ∼1.5–3.5 ng/mL, the correlation between the change in cardiac output (ΔCO) (from prepregnancy values) and the change in plasma relaxin (ΔRLX) (from prepregnancy values—nondetectable) becomes positive. Multiple regression equation: ΔCO = 2610.868 − 2.339 × ΔRLX − 0.458 × Δestradiol + 0.001 × ΔRLX × Δestradiol; P values for ΔRLX, Δestradiol, and ΔRLX × Δestradiol were <0.001, <0.028 and <0.031, respectively. Overall R² = 0.245, P < 0.001.

Relaxin and systemic vascular resistance.

There were significant positive correlations between plasma relaxin concentration and systemic vascular resistance (SVR) especially at 5–6 gestational wk among the women who conceived spontaneously (1 CL; Table 5, Fig. 6B), which were consistent with the inverse relationships between relaxin and CO as described above. These correlations were relaxin vs. SVR: r = 0.761, r² = 0.579, P < 0.001; log relaxin vs. SVR: r = 0.669, r² = 0.447, P = 0.001; Δrelaxin vs. ΔSVR prepregnant to 5–6 wk: r = 0.627, r² = 0.393, P = 0.002; Δlog relaxin vs. ΔSVR prepregnant to 5–6 wk: r = 0.478, r² = 0.228, P = 0.010. Thus in the spontaneous conception cohort, on average ∼41% of the variability in SVR could be explained by the plasma relaxin concentration. For the women who conceived with controlled ovarian stimulation (>1 CL), the only significant relationship was observed at 10–12 gestational wk for relaxin versus SVR: r = 0.518, r² = 0.268, P = 0.028.

Table 5.

Correlation between plasma relaxin concentration and systemic vascular resistance

Comparison	Group	N	r	r 2	P Value
Spontaneous conceptions (1 CL): 5–6 gestational wk
RLN vs. SVR	1 CL	21	0.761	0.579	0.000
Log RLN vs. SVR	1 CL	21	0.669	0.447	0.001
RLN vs. ΔSVR	1 CL	21	0.627	0.393	0.002
Log RLN vs. ΔSVR	1 CL	21	0.478	0.228	0.010
Spontaneous conceptions (1 CL): 7–9 gestational wk
RLN vs. SVR	1 CL	21	0.355	0.126	0.114
Log RLN vs. SVR	1 CL	21	0.394	0.156	0.077
RLN vs. ΔSVR	1 CL	21	0.203	0.041	0.377
Log RLN vs. ΔSVR	1 CL	21	0.206	0.042	0.371
Spontaneous conceptions (1 CL): 10–12 gestational wk
RLN vs. SVR	1 CL	21	0.484	0.234	0.026
Log RLN vs. SVR	1 CL	21	0.491	0.241	0.024
RLN vs. ΔSVR	1 CL	21	0.336	0.113	0.136
Log RLN vs. ΔSVR	1 CL	21	0.336	0.113	0.137
Spontaneous conceptions (1 CL): 14–16 gestational wk
RLN vs. SVR	1 CL	20	0.194	0.038	0.413
Log RLN vs. SVR	1 CL	20	-0.029	0.001	0.903
RLN vs. ΔSVR	1 CL	20	0.096	0.009	0.689
Log RLN vs. ΔSVR	1 CL	20	-0.040	0.002	0.866
Spontaneous conceptions (1 CL): 23–25 gestational wk
RLN vs. SVR	1 CL	20	0.401	0.161	0.080
Log RLN vs. SVR	1 CL	20	0.344	0.119	0.137
RLN vs. ΔSVR	1 CL	20	0.470	0.221	0.036
Log RLN vs. ΔSVR	1 CL	20	0.480	0.230	0.032
Spontaneous conceptions (1 CL): 32–35 gestational wk
RLN vs. SVR	1 CL	19	0.460	0.212	0.047
Log RLN vs. SVR	1 CL	19	0.407	0.165	0.084
RLN vs. ΔSVR	1 CL	19	0.555	0.308	0.014
Log RLN vs. ΔSVR	1 CL	19	0.466	0.217	0.044

Because prepregnant plasma relaxin (RLN) concentration was undetectable (<7.8 pg/mL), the plasma RLN concentration measured at 5–6 gestational wk was used. CL, corpus luteum; SVR systemic vascular resistance. Boldface indicates significance. N, No. subjects.

Relaxin, sFLT1, and sFLT1-to-PLGF ratio.

Surprisingly, there were also significant, albeit modest, positive correlations between plasma relaxin and sFLT concentrations among the women who conceived spontaneously (1 CL) during the first trimester, for example, at 5–6 gestational wk (relaxin vs. sFLT1: r = 0.436, r² = 0.190, P = 0.048; log relaxin vs. sFLT1: r = 0.476, r² = 0.226, P = 0.029; log relaxin vs. ΔsFLT1 prepregnant to 5–6 wk: r = 0.452, r² = 0.204, P = 0.040) (Fig. 6C). Over the entire first trimester, plasma relaxin versus sFLT1 concentrations yielded r = 0.659, r² = 0.434, P < 0.001. There were no significant relationships between plasma relaxin and sFLT1 concentrations for the women who conceived with controlled ovarian stimulation (>1 CL) (data not shown). Plasma relaxin concentration and sFLT1-to-placental growth factor (PLGF) ratio (sFLT1/PLGF) also demonstrated modest, but significant, positive correlations in the spontaneous conception cohort over the first trimester (relaxin vs. sFLT1/PLGF: r = 0.399, r² = 0.159, P = 0.001; log relaxin vs. sFLT1/PLGF: r = 0.426, r² = 0.181, P < 0.001).

sFLT1, sFLT1/PLGF, cardiac output, and SVR.

Because plasma relaxin and sFLT1 concentrations (or sFLT1/PLGF) showed a positive, albeit modest, association (vide supra), we considered the possibility that sFLT1 may have contributed to the reduction in CO and augmentation of SVR, in the face of higher plasma relaxin concentrations. However, the correlations between plasma sFLT1 concentration (or sFLT1/PLGF) and CO, though negative, were generally not significant for any of the three participant cohorts at 5–6, 7–9, or 10–12 wk of gestation or when these gestational ages were combined. The occasional exceptions were modest, e.g., sFLT1 versus CO for spontaneous pregnancies at 5–6 wk (r = −0.424, r² = 0.180, P = 0.056) and 10–12 wk (r = −0.432, r² = 0.186, P = 0.051). Similarly, the correlations between plasma sFLT1 concentration (or sFLT1/PLGF) and SVR, though positive, did not reach statistical significance, except for log sFLT1/PLGF versus log SVR for spontaneous pregnancies and controlled ovarian stimulation combined across the first trimester (r = 0.182, r² = 0.033, P = 0.05). Taken together, the data indicated a weak relationship at best between plasma sFLT1 concentration (or sFLT/PLGF) and CO or SVR during the first trimester of pregnancy irrespective of the mode of conception.

PLGF, cardiac output, and SVR.

There were no significant correlations between plasma PLGF or log PLGF concentrations and CO during the first trimester. However, there were significant, albeit modest, positive associations after the first trimester (gestational weeks 14–16, 23–25, and 32–35), which were more pronounced in women who conceived using programmed cycles (0 CL). The Pearson r, r², and associated P values ranged from r = 0.181, r² = 0.033, and P = 0.094 to r = 0.263, r² = 0.069, and P = 0.01 for the spontaneous conceptions and controlled ovarian stimulation cohorts, after the first trimester. In contrast, the Pearson r and r² values ranged from r = 0.436 and r² = 0.190 to r = 0.482 and r² = 0.232, all being P < 0.001, for the women who conceived using programmed cycles after the first trimester. Both plasma PLGF and log PLGF concentrations also showed significant, negative associations with SVR (r = −0.349, r² = 0.122, P = 0.006 and r = −0.377, r² = 0.142, P = 0.003, respectively) for the women who conceived using programmed cycles. On balance, plasma PLGF concentration accounted for ∼20% and 15% of the variance in CO and SVR, respectively, after the first trimester in the women who conceived using programmed cycles. Although modest, the results stand in contrast to those observed in the other two participant cohorts, in which only 3–7% of the variance in CO and SVR could be explained by plasma PLGF concentration.

Relationship between other hormones and cardiac output.

There was little, if any, association between plasma progesterone, estradiol, hCG, and cortisol concentrations or plasma renin activity and CO during the first trimester for each of the three participant cohorts. The highest correlations were at best modest (e.g., programmed cycles, estradiol vs. CO: r = 0.293, r² = 0.086, P = 0.005; spontaneous pregnancies, log progesterone vs. CO: r = 0.214, r² = 0.046, P = 0.049; controlled ovarian stimulation, log estradiol vs. CO: r = 0.310, r² = 0.096, P = 0.008 and log cortisol vs. CO: r = 0.323, r² = 0.104, P = 0.006). Combining the three cohorts, we observed log progesterone vs. CO: r = 0.155, r² = 0.015, P = 0.024; estradiol vs. CO: r = 0.160, r² = 0.026, P = 0.012; and log estradiol vs. CO: r = 0.226, r² = 0.051, P < 0.001.

DISCUSSION

There were several noteworthy findings and conclusions that can be drawn from this work, many of which were unanticipated. 1) Gestational declines in plasma osmolality and sodium concentration were comparable between women who conceived by IVF using programmed cycles without a CL and those who conceived spontaneously with one CL, suggesting that gestational changes of osmoregulation were not dependent upon the CL or circulating relaxin. Consistent with this finding, 2) plasma relaxin (and hCG) concentration was not significantly correlated with plasma osmolality or sodium concentration during the first trimester in women who conceived spontaneously; however, plasma progesterone, estradiol, PRA, and aldosterone each showed significant inverse correlations with plasma osmolality and less strongly with sodium concentration. 3) The gestational decline in plasma uric acid was attenuated in women who conceived by IVF, especially programmed cycles, and this attenuation was at least partly attributable to reduced gestational rise in renal uric acid clearance. 4) Plasma relaxin concentration paradoxically demonstrated significant negative and positive correlations, respectively, with cardiac output and systemic vascular resistance in women who conceived spontaneously, particularly at 5–6 gestational wk; nevertheless, plasma relaxin concentration in the lower range was associated with higher CO and lower SVR, consistent with an intermediary role for relaxin in the systemic vasodilation of early human pregnancy. 5) Unexpectedly, plasma relaxin concentration showed a significant positive correlation with plasma sFLT1 concentration (and sFLT/PLGF) in the women who conceived spontaneously at 5–6 gestational wk, suggesting relaxin stimulation of a counterregulatory mechanism, but plasma sFLT1 concentration (and sFLT1/PLGF) correlated only weakly with CO and SVR. 6) Plasma PLGF concentration demonstrated significant positive and negative correlations with CO and SVR after the first trimester, which were notably stronger for women who conceived using programmed cycles, suggesting a role for PLGF in rescuing the maternal circulation in this IVF cohort after the first trimester. Finally, 7) there was little, if any association between plasma progesterone, estradiol, hCG, cortisol, or plasma renin activity and CO during the first trimester in the three participant cohorts whether analyzed separately or together.

Plasma Osmolality and Sodium Concentrations

Pregnancy is associated with a lower osmotic threshold for release of arginine vasopressin (AVP) and stimulation of thirst. Working together, these physiological changes lead to chronic reductions of plasma osmolality and sodium concentration in gravid rats and women (reviewed in Ref. 12). Consistent with the literature, the present study showed that plasma osmolality decreased during the first trimester by ∼10 mosmol/kgH₂O (Fig. 1, Supplemental Table S4). Correspondingly, plasma sodium also fell by ∼5 mEq/L, which was reassuring (Fig. 2, Supplemental Table S4). However, inexplicably, plasma osmolality and sodium concentration themselves were not as well correlated as we anticipated.

Unexpectedly, plasma osmolality and sodium concentration fell during the first trimester in women who conceived using programmed cycles without a CL and circulating relaxin (Figs. 1 and 2, Supplemental Table S4). In fact, the pattern of decline was comparable to that in the women who conceived spontaneously with 1 CL and physiological concentrations of circulating relaxin (11). Moreover, neither plasma osmolality nor sodium concentration was significantly correlated with relaxin (or log relaxin) concentration during the first trimester pregnancy of women who conceived spontaneously or by IVF using controlled ovarian stimulation (>1 CL) (Tables 2 and 3, Supplemental Tables S5 and S6). Interestingly, however, the decline in plasma sodium was significantly greater in the >1 CL versus 1 CL + 0 CL cohorts (Fig. 2B), which was not reflected by the decline of measured plasma osmolality (or at least significantly so; Fig. 1B). Therefore, we also calculated plasma osmolality with the equation 2 × Na⁺ + 2 × K⁺ + glucose/18 + urea/2.8. Perhaps not unexpectedly, because the calculated plasma osmolality was mainly driven by 2 × Na⁺, the patterns of change in calculated plasma osmolality over pregnancy among the three participant cohorts now matched plasma sodium concentration, including for the >1 CL group (data not shown). Nevertheless, the possibility remains that, in the case of the >1 CL group, in which many of the women have supraphysiological levels of circulating relaxin, the hormone may have contributed to further decline in plasma sodium and (calculated) osmolality in this participant cohort. However, a caveat is that plasma concentrations of progesterone and aldosterone, as well as plasma renin activity (PRA) were also higher during the first trimester in the >1 CL cohort (11) (see below).

Although our previously published pilot study employing the same experimental approach was supportive of a role for relaxin (9) and motivated the more comprehensive and larger investigation reported here, unfortunately the present results did not corroborate this earlier work. Possible explanations for the discrepancy are that historical controls were used as a comparison group and the decline in plasma osmolality, although attenuated, was not totally inhibited in the earlier study (9). The present study is also at odds with work we recently conducted in another patient population (10). One potential explanation for this inconsistency is that plasma sodium concentration was only assessed once and at an earlier gestational time point (4 gestational wk), whereas in the present longitudinal investigation the earliest pregnancy plasma sample was obtained at 5–6 gestational wk. A more likely reason, however, relates to the different study designs. That is, the present study was longitudinal and thus included a prepregnancy (and postpartum) time point for comparison, in addition to six pregnancy time points. When compared to the prepregnancy baseline, it is clear that plasma osmolality and sodium concentration decreased during the first trimester of pregnancy in women who conceived without a CL or circulating relaxin in a similar fashion to spontaneously conceived pregnancies (1 CL) (Figs. 1 and 2, Supplemental Table S4).

As previously reported, an osmoregulatory effect of administered hCG was noted in intact but not ovariectomized female nonpregnant rats, which suggested the intermediary role of an ovarian hormone like relaxin (13). Indeed, the evidence for a role of relaxin in the osmoregulatory changes of pregnancy in the conscious, gravid rat model is compelling (reviewed in Ref. 12). The lack of an hCG effect in men, despite an apparent osmoregulatory role when administered to women in the luteal phase, also suggested that there may be an intermediary role for an ovarian hormone in women, too (13). However, neither relaxin nor hCG was implicated in the gestational decline of plasma osmolality or sodium concentration in the present study (Tables 2 and 3, Supplemental Tables S5 and S6).

Interestingly, plasma osmolality (and to a lesser degree sodium concentration) significantly correlated with plasma estradiol and progesterone concentrations, which were themselves highly correlated as expected (Fig. 3, Tables 2 and 3, Supplemental Tables S5 and S6). These observations are consistent with the meticulous investigations conducted in nonpregnant women in which sex steroids, particularly estradiol, were implicated in decreasing the plasma osmolality and osmotic threshold of AVP release during the luteal phase and after administration of oral contraceptives (14–16). Thus, estradiol may be a primary hormone underlying the osmoregulatory changes of pregnancy, too.

In addition, there were modest, but significant, inverse correlations between plasma osmolality and PRA and between plasma osmolality and aldosterone concentration in all three cohorts (Tables 2 and 3). The correlations of PRA and aldosterone with sodium concentration were considerably weaker (Supplemental Tables S5 and S6). Circulating angiotensin II was reported to stimulate AVP secretion and increase thirst (3). Furthermore, angiotensin II was shown to regulate renal water reabsorption by direct renal actions including targeting of AQP2 to the apical membrane in the collecting duct (9, 17). These actions of angiotensin II were consistent with the finding of increased urinary excretion of AQP2 in pregnant women and increased expression of AQP2 mRNA and protein in the renal papillae of gravid rats (2, 17, 18). Administration of angiotensin II to reach concentrations that were nondipsogenic potentiated the dipsogenic action of infused relaxin in rats. Central administration of an AT₁ antagonist blocked this dipsogenic action of circulating relaxin (19). Therefore, circulating angiotensin II may conceivably contribute to the osmoregulatory changes of pregnancy either in concert with or independently of relaxin. However, the inverse correlation between plasma osmolality and PRA was also observed in women without circulating relaxin in the present study (programmed cycles, 0 CL; Tables 2 and 3). Thus, if circulating angiotensin II contributed to the osmoregulatory changes of pregnancy, it apparently did so independently of relaxin. With respect to aldosterone, it was reported to increase urinary excretion of water and promote hypernatremia, i.e., changes opposite to those observed during pregnancy (17, 20). Finally, it is possible that the associations of PRA and aldosterone concentration with plasma osmolality and, more weakly, sodium concentration were correlational but not causal and arose because PRA and aldosterone concentration correlated significantly with plasma progesterone concentrations in these women (11).

We conclude that, in contrast to gravid rats, circulating relaxin does not appear to be necessary for gestational osmoregulatory changes in women. Plasma osmolality concentrations fell during the first trimester in the women who conceived using programmed cycles without a CL and without detectable circulating relaxin (<7.8 pg/mL; Ref. 11). This decline was statistically indistinguishable from the two cohorts with circulating relaxin (1 CL and >1 CL). Rather, our work supported a role for estradiol and/or progesterone, as well as possibly angiotensin II. It should be noted that estradiol and progesterone are replaced for the first 10 wk of gestation in women who conceive using programmed cycles and who do not develop a CL.

Plasma Uric Acid and Renal Uric Acid Clearance

Plasma uric acid concentration declines during the first trimester of normal human pregnancy because renal uric acid clearance rises due to of increased glomerular filtration rate (GFR) and hence increased filtered load of uric acid, reduced tubular reabsorption, or both (reviewed in Ref. 21). The present study is consistent with the literature showing reciprocal changes in plasma uric acid concentration and renal uric acid clearance in the cohort who conceived spontaneously (Figs. 4 and 5, Supplemental Table S4). However, plasma uric acid declined to a lesser degree during early pregnancy in the women who conceived by IVF, especially using programmed cycles. This was mirrored by a delayed and subdued rise in renal uric acid clearance, which likely contributed to the higher plasma uric acid concentration. Unfortunately, our measurements of creatinine clearance were not reliable, so we could not discern whether the attenuated rise in renal uric acid clearance in the IVF cohorts was due to subdued gestational increase in GFR, enhanced tubular reabsorption of uric acid, or both. Of additional note, pathologically high plasma uric acid concentrations were observed during late pregnancy in the women who conceived by programmed cycles (0 CL), which persisted even when the adverse pregnancy outcomes were excluded from the analysis (12). Whether the elevated plasma levels of uric acid were solely due to reduced renal clearance was not clear. Increased uric acid production is another potential causal factor (21).

Microalbuminuria

Microalbuminuria is recognized as a marker of vascular dysfunction and predictor of future cardiovascular events (22). We assessed microalbuminuria in 24-h urine collections before pregnancy, during gestational weeks 5–6, when the kidneys were presumably hyperfiltrating, and 6 mo after pregnancy. Overall, microalbuminuria was more frequent in the subjects who conceived by IVF compared with spontaneously conceived pregnancies (40% vs. 20%)(Supplemental Table S7). However, we did not see a difference in frequency or magnitude of detectable microalbuminuria after pregnancy between women who conceived spontaneously and by IVF, indicating no long-term harmful impact of IVF on vascular function. When detected, the degree of microalbuminuria was within the normal limits in 50% of cases (0–29 mg/g creatinine) and ranged from 31 to 164 mg/g creatinine in the other 50%. Our study of microalbuminuria was admittedly small and exploratory in scope; as such, it needs repeating in larger cohorts of women who conceive spontaneously or by IVF.

Circulating Hormones, Cardiac Output, and Systemic Vascular Resistance

Administration of relaxin to conscious, chronically instrumented male and nonpregnant female rats significantly increased CO and decreased SVR by ∼25% (13, 23, 24). In addition, administration of rat relaxin-neutralizing antibodies abrogated the gestational rise in CO and fall in SVR in the conscious gravid rat model, at least during midterm pregnancy (25). These findings motivated our recent work, in which we utilized IVF as a means to determine whether relaxin might exert a similar intermediary role in human pregnancy. Indeed, women who conceived using programmed cycles without a CL and circulating relaxin demonstrated a delayed and attenuated rise in CO and decline in SVR during the first trimester relative to women who conceived spontaneously or by controlled ovarian stimulation, who were comparable (7). Therefore, in the present study we reasoned that there should be (strong) positive and negative correlations, respectively, between plasma relaxin concentration and CO and between plasma relaxin concentration and SVR.

Surprisingly, we observed the opposite relationships. Plasma relaxin concentration and CO were inversely related and plasma relaxin concentration and SVR directly correlated (Tables 4 and 5). These associations were strong and statistically significant at many time points during the first trimester in women who conceived spontaneously. Approximately 40% of the variability in CO and SVR could be attributed to plasma relaxin concentration. The scatterplots depicting plasma relaxin concentration versus change in CO and SVR from prepregnant baseline at 5–6 gestational wk for women who conceived spontaneously strongly reinforced these unanticipated inverse relationships (Fig. 6, A and B). Indeed, greater gestational decreases and increases in SVR and CO, respectively, were associated with smaller gestational rises of plasma relaxin concentration during the first trimester. One possible interpretation is that plasma relaxin concentration in the lower range contributed to maternal systemic vasodilation and increased CO during the first trimester. This explanation is consistent with our earlier work in which the maternal systemic hemodynamic changes of early pregnancy were markedly attenuated in women without circulating relaxin (i.e., programmed cycles and 0 CL) (7). Of course, we cannot definitively conclude that relaxin is the critical CL hormone in human pregnancy as we were able to do in the gravid rat model by using specific relaxin-neutralizing antibodies (25). The next best approach in women might eventually be to replace relaxin in those who conceived by programmed cycles without circulating relaxin, in order to determine whether the relaxin infusion would restore CO and SVR to the physiological levels normally observed in the first trimester (and reduce preeclampsia risk). Conceivably, however, another yet to be identified CL hormone(s) might be the important vasodilator of early pregnancy in women.

As stated above, lower plasma relaxin concentrations were paradoxically associated with higher CO and lower SVR, and higher plasma relaxin concentrations were associated with lower CO and higher SVR (Fig. 6. A and B, Tables 4 and 5). Thus, plasma relaxin concentration in the lower range may have contributed to maternal systemic vasodilation and increased CO during the first trimester, although higher relaxin concentrations could have promoted coupling of RXFP1 with different G proteins, downregulated RXFP1 through a β-arrestin-mediated or another mechanism, and/or upregulated countervailing vasoconstrictor pathways, thereby opposing its own vasodilatory effect (17, 18, 26). Teleologically speaking, by limiting its own potent vasodilatory action, relaxin may prevent undue maternal systemic vasodilation and excessive falls of maternal blood pressure and uterine blood flow during pregnancy. This apparently biphasic action of relaxin is not unprecedented, although the underlying mechanism(s) remains uncertain (13, 27).

We speculated that sFLT1 might be one potential counterregulatory vasoconstrictor that could be stimulated in the maternal vasculature or placenta, especially by higher plasma concentrations of relaxin. Indeed, we found a significant positive relationship between plasma relaxin and sFLT1 concentrations (or sFLT1/PLGF) for many of the time points during the first trimester in women who conceived spontaneously (Fig. 6C; see results). However, the relationships between sFLT1 (or sFLT1/PLGF) and CO or SVR, although in the expected direction, were modest at best and of borderline significance. Therefore, even though plasma relaxin and sFLT1 concentrations were directly and significantly correlated, sFLT1 did not appear to be a mechanism for restraining the vasodilatory action of higher plasma relaxin concentrations.

In addition to the conclusion that plasma relaxin concentration in the lower range may have contributed to the gestational increase in CO during the first trimester, there is an alternative, but not mutually exclusive, possibility that circulating relaxin and estradiol cooperate. First, both hormones are vasodilators and share cellular mechanisms of vasodilation including through nitric oxide (20, 28). Second, multiple linear regression analysis suggested an interaction between these two hormones such that the unexpected negative relationship between plasma relaxin and CO became a positive association when plasma estradiol concentration exceeded ∼2.5 ng/mL (Fig. 7).

Our original hypothesis was that relaxin contributes to maternal systemic vasodilation during early human pregnancy and placental factors supersede after the first trimester, when the organ has sufficiently developed (6). Indeed, PLGF and CO demonstrated a significant positive correlation and PLGF and SVR a significant negative correlation, but only after the first trimester, when plasma PLGF was starting to rise precipitously (11). These correlations were notably stronger for women who conceived using programmed cycles lacking a CL and circulating relaxin but again only after the first trimester. Thus, PLGF (and perhaps other placental vasodilators) may have rescued the maternal systemic circulation in women who conceived using programmed cycles, because the gestational changes in CO and SVR recovered after the first trimester in this participant cohort (7). Although this participant cohort had a significantly higher risk for developing preeclampsia and preeclampsia with severe features, the majority did not develop the disorder (2). Perhaps by restoring maternal vasodilation after the first trimester, PLGF prevented preeclampsia risk from being even higher in this IVF cohort.

Perspectives and Significance

Many findings that arose from this investigation were unexpected, underscoring the complexity of pregnancy biology. Indeed, this study perhaps raised as many or more questions than it provided answers. Contrary to our expectation, relaxin was not necessary for the reductions in plasma osmolality and sodium concentration during the first trimester, nor did the hormone correlate with these variables. Rather, modest but significant correlations were observed between plasma osmolality and estradiol or progesterone, suggesting that, perhaps not unlike the nonpregnant condition, sex steroids also modulate osmoregulation during pregnancy. In a similar vein, rather than observing a positive correlation between plasma relaxin concentration and CO and a negative relationship with SVR, we found the opposite. Nonetheless, these paradoxical correlations are not inconsistent with a role for relaxin in the maternal changes of systemic hemodynamics in the first trimester; it is just that lower plasma relaxin concentrations appeared to be more effective. This interpretation is consistent with earlier work in which the changes of systemic hemodynamics in early pregnancy were attenuated in women without circulating relaxin (i.e., programmed cycles and 0 CL) (7). However, we cannot exclude the potential role of other, as of yet unidentified CL factors in the regulation of maternal vasodilation during early pregnancy. Another possible explanation is that there may be cooperation among two or more hormones whose combined actions mediate gestational decreases in SVR and increases in CO. Indeed, evidence for such cooperation between relaxin and estradiol was noted in this study. An additional unexpected finding was that plasma relaxin and sFLT1 concentrations were directly correlated, raising the possibility that higher plasma relaxin stimulated more sFLT1 production, thus restraining the vasodilatory action of relaxin and preventing undue maternal systemic vasodilation. However, we did not find significant associations between sFLT1 or sFLT1/PLGF and CO or SVR. Rather, higher plasma relaxin concentrations may have stimulated other countervailing vasoconstrictors besides sFLT1 or perhaps partially desensitized the relaxin receptor RXFP1 during early gestation. If so, then this raises the interesting possibility that the failure of relaxin administration to improve acute heart failure in the recently completed phase IIIA clinical trial may conceivably be explained by excessive circulating relaxin concentrations attained in that study, on average ∼10 ng/mL (29, 30).

SUPPLEMENTAL DATA

Supplemental Tables S1–S7: https://doi.org/10.6084/m9.figshare.15127500.

GRANTS

This work was supported by P01 HD065647-01A1 from the National Institute of Child Health and Human Development (PD/PI K.P.C), the J. Robert and Mary Cade Professorship of Physiology (K.P.C), the Gatorade Trust through funds distributed by the University of Florida College of Medicine and the Division of Nephrology, Hypertension, and Renal Transplantation, Department of Medicine (M.S.S), and matching funds from the University of Florida College of Medicine (K.P.C). Research reported in this publication was also supported by the University of Florida Clinical and Translational Science Institute, which is underwritten in part by the NIH National Center for Advancing Translational Sciences under award number UL1TR001427. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Study data were collected and managed with REDCap electronic data capture tools hosted at the University of Florida.

DISCLOSURES

K.P.C. discloses use patents for relaxin. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

K.P.C. and M.S.S. conceived and designed research; M. Lingis performed experiments; K.P.C., S.T., Y.-Y.C., Y.Q., M. Li and M.K.-W. analyzed data; K.P.C. and M.K.-W. interpreted results of experiments; K.P.C. and Y.Q. prepared figures; K.P.C., S.T., Y.-Y.C. and M. Lingis drafted manuscript; K.P.C., S.T., Y.-Y.C., Y.Q., M. Li, M. Lingis, R.S.W., A.R.-V., M.K.-W., and M.S.S. edited and revised manuscript; K.P.C., S.T., Y.-Y.C., Y.Q., M. Li, M. Lingis, R.S.W., A.R.-V., M.K.-W., and M.S.S. approved final version of manuscript.