Correlation of fetal heart rate dynamics to inflammatory markers and brain-derived neurotrophic factor during pregnancy

Abstract

Objectives

This study aims to show the relation between biomarkers in maternal and cord-blood samples and fetal heart rate variability (fHRV) metrics through a non-invasive fetal magnetocardiography (fMCG) technique.

Methods

Twenty-three women were enrolled for collection of maternal serum and fMCG tracings immediately prior to their scheduled cesarean delivery. The umbilical cord blood was collected for measurement of biomarker levels. The fMCG metrics were then correlated to the biomarker levels from the maternal serum and cord blood.

Results

Brain-derived neurotrophic factor (BDNF) had a moderate correlation with fetal parasympathetic activity (0.416) and fetal sympathovagal ratios (−0.309; −0.356). Interleukin (IL)-6 also had moderate-sized correlations but with an inverse relationship as compared to BDNF. These correlations were primarily in cord-blood samples and not in the maternal blood.

Conclusions

In this small sample-sized exploratory study, we observed a moderate correlation between fHRV and cord-blood BDNF and IL-6 immediately preceding scheduled cesarean delivery at term. These findings need to be validated in a larger population.

Introduction

Numerous biomarkers have been studied during gestation, where inflammatory processes and neurotrophic factors have been of particular interest in the past decade [1], [2], [3]. Neurotrophins contribute to the growth, development, maturation, and protection of neurons of the central and peripheral nervous system [4], [5], [6], [7]. Synthesized at low levels in the central nervous system (CNS), brain-derived neurotrophic factor (BDNF) is a neurotrophin that is involved during early embryonic development [6], [7], [8]. During gestation, neurotrophins are said to have different simultaneous origins, present in maternal, placental, fetal, and amniotic fluid compartments [8]. The origin of cord-blood BDNF remains unclear [5, 9]. It has been speculated that cord-blood BDNF may be derived from the BDNF secreted from the placenta [5].

Pregnancy can be characterized by elevations of both anti- and pro-inflammatory cytokine levels as an immune response [10]. Maternal inflammation or an inflammatory intrauterine environment have been reported to affect fetal development [10, 11]. BDNF takes part in a parallel activation of anti-inflammatory mechanisms [6]. BDNF, under certain scenarios, is said to suppress pro-inflammatory cytokines such as tumor necrosis factor (TNF)-α and interleukin (IL)-6 [6, 12, 13]. Reducing IL-6 levels in the liver downregulates the acute-phase-reactant C-reactive protein (CRP) [12]. Both molecules, i.e., IL-6 and CRP, are frequently used to assess the presence of inflammation and are easy to identify in serum [14]. The relative importance of the functional roles for inflammatory markers throughout gestation remains to be fully explained. However, effects of maternal inflammation have been associated with fetal development, including the risk of neurobehavioral disorders [15, 16].

During fetal life, the autonomic nervous system (ANS) undergoes a process of maturation and development [17]. The ANS development and its functioning has been reported to correlate to heart rate variability (HRV) [17]. For this reason, many studies have tried to establish correlations between changes in fetal HRV (fHRV) and fetal development [17]. BDNF is said to contribute to the development of the CNS [7], and also may have a role in regulating heart rate in adults [18]. As for IL-6 and CRP, both have been reported to be inversely correlated to vagal (parasympathetic) nerve activity assessed by HRV on adults [19, 20]. However, Spann et al. 2018 showed that only maternal CRP levels were inversely correlated to fHRV [11].

To measure fHRV, researchers have relied on non-invasive biomagnetic recordings (fetal magnetocardiography – fMCG), which allow a more stable and precise HRV assessment than the standard approach of cardiotocography [17, 21]. MCG is a magnetic homolog of electrocardiography (ECG) and is well suited for fetal recording since it is completely non-invasive. However, to our knowledge there are no reported human fetal studies that address the association between the fHRV and inflammatory markers, neurotrophic factors, and other maturation molecules. In this study, we explored the correlation between fetal-development biomarkers in maternal and cord blood samples with measures of fHRV obtained by means of fMCG.

Materials and methods

Participants and samples collection

Subjects eligible included English and Spanish speaking pregnant women between the ages of 18–40 years who received medical care at the university hospital. The focus of this study was a convenience sample where subjects were included only if they were scheduled for a cesarean delivery close to term based on the electronic chart review. The study population included both low-risk and high-risk subjects with exclusions for: fetal malformations, genetic anomalies and/or multiple pregnancies. The study was approved by the Institutional Review Board (Protocol Numbers: #04234 on June 14th, 2017 and # 206700 on June 5th, 2017. All the subjects provided informed written consent.

Subject demographics and other pregnancy-related characteristics are provided in Table 1. As seen from the table, there is a mix of both low-risk and high-risk subjects who participated in the study, out of which the predominant high-risk condition was pregestational diabetes, either Type 1 or 2 (12 out of 23). Maternal subjects participated in the fMCG study immediately prior to their scheduled cesarean delivery with estimated GA ranging from 36 weeks/6 days to 39 weeks/3 days.

Table 1:

Participant demographics and conditions.

Subject	GA at recording	Maternal age	Gravid	Parity	Apgar 1 min	Apgar 5 min	Maternal origin	Ethnicity	Conditions
01	39w3d	23	4	1	8	9	White	Non-H/L	N/A
02	39w2d	29	1	0	8	9	White	Non-H/L	N/A
03	39w2d	35	2	1	9	9	Black	Non-H/L	Type 2 diabetes, CHTN, obesity
04	38w3d	38	2	1	8	9	Asian	H/L	Type 2 diabetes, CHTN, hx of renal failure
05	37w2d	36	4	1	8	9	White	H/L	Type 1 diabetes, obesity, seizures, cocaine abuse
06	39w1d	36	4	3	7	8	White	Non-H/L	Type 2 diabetes, obesity, CHTN, +HPV
07	38w2d	27	3	1	8	9	White	H/L	Type 1 diabetes, GHTN versus CHTN
08	37w2d	35	4	2	7	8	Black	Non-H/L	Type 2 diabetes, CHTN
09	36w6d	25	2	1	7	8	White	Non-H/L	Type 1 diabetes
10	38w1d	32	3	2	8	9	Black	Non-H/L	Hx preeclampsia
11	39w0d	30	4	2	8	9	Black	Non-H/L	N/A
12	39w2d	22	2	1	8	9	White	Non-H/L	N/A
13	39w0d	25	4	3	3	8	Black	Non-H/L	N/A
14	38w5d	40	2	1	8	9	Black	Non-H/L	CHTN, obesity, sickle cell trait,
15	38w2d	32	3	1	8	9	White	Non-H/L	Type 1 diabetes, hypothyroidism, obesity, asthma
16	39w1d	27	4	1	1	7	Black	Non-H/L	Type 2 diabetes, obesity
17	38w5d	28	2	0	8	9	White	Non-H/L	Type 1 diabetes
18	39w1d	36	3	2	8	9	White	Non-H/L	N/A
19	39w0d	31	3	1	8	9	White	Non-H/L	N/A
20	37w0d	30	2	1	8	9	White	Non-H/L	Type 1 diabetes, GHTN
21	38w0d	32	2	1	8	9	Black	Non-H/L	Type 2 diabetes, hypothyroid, benign, essential hypertension, obesity, SVT
22	39w0d	35	2	1	9	9	White	Non-H/L	N/A
23	38w5d	31	2	1	1	8	Black	Non-H/L	Obesity, hypothyroidism

H/L, Hispanic or Latino; Non-H/L, non-Hispanic or Latino.

The maternal blood was collected immediately prior to delivery for measurement of maternal serum biomarkers: BDNF, IL-6, and CRP. The cord blood was collected at delivery for measurement of the same biomarkers, to be representative of the fetus’s inflammatory biomarkers. Blood was collected into sodium heparin tubes and placed immediately on ice. Plasma was collected after centrifuging the blood for 15 min at 2000×g at 4 °C. Plasma was frozen and stored at −80 °C.

CRP was measured in plasma (diluted 1:10000 in assay buffer) by ELISA (Cayman Chemical, Ann Arbor, MI, USA) following the manufacturer’s protocol. BDNF and IL-6 were measured in plasma using the MILLIPLEX^® MAP Human Myokine Magnetic Bead Panel (with only BDNF and IL-6 analytes) following the manufacturer’s instructions, and data was acquired on a Luminex^® 200 with xPONENT^® software.

fMCG recordings

Biomagnetic signals were recorded using a SQUID-based non-invasive 151-sensor SARA (SQUID Array for Reproductive Assessment) system [22]. Fetal MCG was extracted as a subset of biomagnetic recordings obtained from the fetus. The gestational age during the recording ranged from 36 to 39 weeks. One recording (limited to 10 min) from each of the 23 pregnant mothers was included in the analysis. Although the protocol was set for a 20-min recording session, some of the subjects did not complete the whole session as they became either uncomfortable or exhibited an increase in maternal-motion artifacts towards the later part of the recording. In order to be consistent across subjects we included only the first 10 min of the recording in the analysis.

Data analysis

From the raw biomagnetic recordings, fMCG tracings were derived and analyzed after the removal of interfering maternal cardiac signals using established methods [23]. FHRV was quantified using the following standard measures: mean heart rate (HR), standard deviation of normal-to-normal beat (SDNN), and root mean square of successive differences (RMSSD). These time-domain measures reflect parasympathetic (vagal) and sympathetic activity. Mean HR and SDNN are said to reflect both types of activities. Meanwhile, as a parameter of short-term variability, RMSSD mainly reflects vagal activity [24]. Frequency-domain measures were derived from the spectrum estimate of the RR intervals divided as follows into very low frequency (VLF, 0.02–0.08 Hz), low frequency (LF, 0.08–0.2 Hz), and high frequency (HF, 0.4–1.7 Hz) bands. Furthermore, the following power ratios were calculated for each fetus: VLF/LF, VLF/HF, and LF/HF. These frequency bands and ratios reflect the fetal sympatho-vagal balance and have been detailed in several fetal heart studies [25–27].

Since fHRV can be affected by fetal state, we quantified the fetal behavioral states in these recordings. We followed the four fetal behavioral states that were defined by Nijhuis et al., which follow a specific combination of eye and body movements, and fHRV patterns [28]. Using fMCG, the fetal body movement can be quantified by means of an actogram. Although fetal eye movements cannot be detected with magnetography, they are not required for a reliable classification of fetal behavioral states [29]. Therefore, the heart signals with high temporal resolution taken from the fMCG along with the fetal movement quantified by the actogram were used to identify the four behavioral states: quiet sleep (1F), active sleep (2F), quiet awake (3F), and active awake (4F). By definition, these behavioral states are only considered as such when they last a minimum of 3 min [28].

Statistical analysis

Correlation analysis using Pearson correlation coefficients was used to examine the fHRV parameters for the strength of their associations with BDNF, IL-6, and CRP. To reduce right-skewing and make the distributions more nearly normal for Pearson correlation analysis, BDNF (in picograms per milliliter), IL-6 (in mean fluorescence-intensity units), and CRP (in nanograms per milliliter) were transformed to their natural logarithms, as were SDNN (in milliseconds), RMSSD (in milliseconds), and the three band ratios (in dimensionless units). As mentioned above, because approximately half of the subjects had a pregestational diabetic condition, we calculated the total correlations as unadjusted for pregestational diabetic status, whereas partial correlations represented the correlations that remain after “partialling out” the effect of pregestational diabetes. Point-biserial correlation coefficients of fHRV parameters with pregestational diabetes status (coded Yes=1, No=0) were also estimated. Cohen’s effect-size criteria [30] were used as thresholds to classify correlation coefficients descriptively by their absolute value (|r|) as follows: |r|≥0.5=large, 0.3≤|r|<0.5=moderate, and 0.1≤|r|<0.3=small, with |r|<0.1 being considered negligible. Statistical null-hypothesis tests were not conducted.

Results

The four behavioral states were quantified for all subjects. In average, the most frequent state was 1F (54.78 %) followed by 2F (28.26 %). State 3F hardly occurred (0.01 %), and only one subject presented the state 4F (2.4 %). The time in which no state was classified made up 14.6 % of the total. Because of the low number of occurrences and short durations of the 3F and 4F states, the study subjects are primarily considered in a fetal sleep state (1F and 2F) for the purpose of fHRV analysis.

Table 2 shows the mean and standard deviation for all the measures including blood biomarkers and fetal heart parameters. Table 3 contains the total and partial Pearson correlation coefficients between the fHRV parameters and the biomarkers of BDNF, IL-6, and CRP from both the maternal and cord blood; the point-biserial correlations with pregestational diabetes status are also shown for completeness. When measured in maternal blood, BDNF, IL-6, and CRP had small-to-negligible total correlations with most fHRV parameters. The only exception was IL-6, which showed a moderate total correlation with mean HR (0.303). Similarly, when adjusting for pregestational diabetes, only BDNF had a moderate partial correlation with SDNN (0.354). However, when measured in cord blood, some markers showed sizeable correlations with multiple fHRV parameters. BDNF had a moderate positive correlation, both total (0.416) and partial (0.393), with RMSSD, which represents the parasympathetic (vagal) system. Regarding the sympathovagal-balance ratios, VLF/HF and LF/HF had moderate negative total correlations with cord-blood BDNF (−0.356 and −0.309, respectively). After controlling for pregestational diabetes status, these two parameters retained their moderate correlations with cord-blood BDNF (−0.409 for both parameters).

Table 2:

Mean and standard deviation for blood biomarkers and fetal heart parameters.

Type of variable	Name of variable	Blood sample
		Maternal		Cord
		Mean	Std dev	Mean	Std dev
Blood biomarkers	log(BDNF)	7.261	0.890	6.751	1.202
	log(IL-6)	3.146	0.237	3.196	0.175
	log(CRP)	8.925	1.610	5.371	1.388
Fetal HR	Mean HR, bpm	143.450	10.979	142.418	10.779
fHRV metric	log(SDNN)	2.817	0.383	2.801	0.378
	log(rMSsd)	2.056	0.402	2.067	0.412
	log(vlf/hf)	0.673	0.996	0.658	1.034
	log(vlf/lf)	0.931	0.440	0.907	0.418
	log(lf/hf)	−0.259	1.027	−0.249	1.076

fHRV, fetal heart rate variability; BDNF, brain-derived neurotrophic factor; IL-6, interleukin-6; CRP, C-reactive protein; HR, heart rate; RMSSD, root mean square of successive differences; SDNN, standard deviation of normal-to-normal beat; VLF, very low frequencies; LF, low frequencies; HF, high frequencies.

Table 3:

Correlation between fetal heart rate variability metrics and biomarkers.

fHRV metric	Correlation	BDNF		IL-6		CRP		Diabetes status
		Maternal	Cord blood	Maternal	Cord blood	Maternal	Cord blood	Maternal	Cord blood
Influenced by sympathetic and parasympathetic systems
Mean HR	Total	0.243	0.169	0.303	0.3	−0.096	0.096	0.391	0.372
	Partial	0.139	0.268	0.241	0.281	−0.055	0.101	–	–
Log SDNN	Total	0.254	0.065	−0.099	−0.131	0.046	0.076	−0.234	−0.386
	Partial	0.354	−0.014	−0.05	−0.097	0.018	0.086	–	–
Log RMSSD	Total	0.03	0.416	−0.098	−0.344	−0.046	−0.081	−0.097	−0.187
	Partial	0.064	0.393	−0.079	−0.331	−0.058	−0.081	–	–
Sympathovagal balance
Log VLF/LF	Total	0.05	−0.122	0.077	−0.17	0.057	0.274	0.487	0.456
	Partial	−0.122	−0.036	−0.036	−0.249	0.132	0.305	–	–
Log VLF/HF	Total	0.144	−0.356	0.064	0.338	0.166	0.133	−0.197	−0.19
	Partial	0.22	−0.409	0.112	0.367	0.146	0.136	–	–
Log LF/HF	Total	0.118	−0.309	0.029	0.389	0.136	0.038	−0.4	−0.34
	Partial	0.278	−0.409	0.131	0.456	0.098	0.043	–	–

Bold numbers indicate moderate correlations (0.3≤|r|<0.5); fHRV, fetal heart rate variability; BDNF, brain-derived neurotrophic factor; IL-6, interleukin-6; CRP, C-reactive protein; HR, heart rate; RMSSD, root mean square of successive differences; SDNN, standard deviation of normal-to-normal beat; VLF, very low frequencies; LF, low frequencies; HF, high frequencies.

Regarding cord-blood inflammatory markers, IL-6 had moderate total correlations with the same fHRV parameters as BDNF, but with an inverse behavior. Specifically, it had total correlations of −0.344 with RMSSD, 0.338 with VLF/HF, and 0.389 with LF/HL. After controlling for pregestational diabetes, IL-6 continued to have moderate partial correlations of −0.331 with RMSSD, 0.367 with VLF/HF, and 0.456 with LF/HF. The only exception was IL-6 in the cord blood, which had a barely moderate (0.300) correlation with mean HR before partialling out pregestational diabetes status, and a small-size (0.281) partial correlation afterward. CRP had only one moderate correlation. It was a partial correlation that was positive with VLF/LF (0.305). The point-biserial correlations between the mean HR and the diabetic status were moderate and positive for both maternal (0.391) and cord blood (0.372). Similarly, VLF/LF had moderate and positive diabetes correlations (0.487 for maternal, and 0.456 for cord blood). However, LF/HF showed negative and moderate diabetes correlations, with the maternal blood having a −0.4, and the cord blood a −0.34. Only the cord blood had a moderate point-biserial diabetes correlation with SDNN (−0.386), since the maternal blood had a small-sized correlation.

Discussion

Our study shows a potential relationship between fetal developmental biomarkers and fHRV parameters. Further, since the main goal of the study was to examine the relation of the biomarkers with the fHRV, we considered the diabetic status as a confounding factor in our analysis, as almost half of our test subjects had pregestational diabetes (either type 1 or type 2).

The cord-blood BDNF levels had a positive correlation with a parameter that reflects parasympathetic modulation (RMSSD) and negative correlations with two parameters of sympathovagal balance. The positive correlation coincides with the study of Wan et al. 2014, which suggests that BDNF enhances parasympathetic activity [18]. Our correlations appear adequate, since altered BDNF levels suggest an abnormal fetal growth and brain development [8]. An altered BDNF level could predispose to cardiovascular disease [7], which in turn would be reflected by a low correlation with the fHRV. Furthermore, it seems that pregestational diabetes had very little influence in controlling for the relationship between BDNF and the fHRV parameters. Although not the same population, it has been observed that the cord-blood BDNF levels are not affected by gestational diabetes [31]. The correlations between BDNF and fHRV were mainly small or negligible on the maternal samples. However, BDNF’s correlation with SDNN increased from 0.254 (small) to 0.354 (moderate) after partialling out pregestational diabetes status, suggesting that confounding with diabetes tends to mask SDNN’s true relationship with BDNF. A similar masking is evident for LF/HF’s correlation with BDNF. Though always small, this correlation increased from 0.118 (close to negligible) before diabetes adjustment to 0.278 (close to moderate) afterward. Again, comparing with respect to maternal gestational diabetes, there have been observations that show a decrease in the expression of BDNF during pregnancy [32]. The small correlations between maternal BDNF and fHRV may probably be because cord-blood BDNF is theorized to come mainly from the placenta [5, 9], although it has been proposed that maternal BDNF possibly supports the development of the fetal CNS by reaching the fetal brain via the utero-placental barrier [6].

Regarding the inflammatory markers, only maternal IL-6 had a moderate correlation with the mean HR, and small or negligible partial correlations when controlling for diabetes. Similarly, maternal CRP only had small or negligible correlations with fHRV parameters, even when controlling for diabetes. Cord-blood IL-6 levels presented moderate correlations with the time-domain parameters of mean HR and RMSSD, and with the sympathovagal ratios VLF/HF and LF/HF. Except for mean HR, these moderate correlations persisted even when controlling for diabetes, possibly due to similar levels of cord-blood IL-6 between control and diabetic mothers. This uniformity of IL-6 levels was reported by Nelson et al. 2007, although it was only studied in offspring of type 1 diabetic mothers [33]. Meanwhile, the cord-blood CRP levels had small or negligible correlations, except for VLF/HF when controlling for diabetes. Even though it has been reported that CRP is higher in cord-blood samples of offspring of type 1 diabetics [33], there was no apparent change for our correlations. Of note, although most correlations of the parameters are low, they are negative for the parasympathetic (vagal) metric and positive for the sympathetic and sympathovagal metrics. These results coincide with several studies in the literature. Thayer et al. 2009 reported that CRP was inversely related with vagally mediated HRV [12]. Al-Shargabi et al. 2017 found an inverse relation between HRV and modulating cytokines [34]. Sloan et al. 2007 assessed vagus-nerve activity by HRV, and found it inversely related to CRP levels [19]. Similarly, Frasure-Smith et al. 2009 found negative correlations between vagus-nerve activity and levels of both IL-6 and CRP [20]. These negative correlations could be due to attenuation of pro-inflammatory cytokine release by vagal activation [35]. Therefore, it is more likely that the increase in BDNF levels for fetal development enhances the parasympathetic activation and thus reduces the levels of IL-6 and CRP, rather than the pro-inflammatory markers diminishing vagal modulation.

The strengths of this study include the non-invasive assessment of fetal heart and ANS development during pregnancy and its relation to neurotrophic factors and inflammatory markers. To our knowledge, there have been no reports of detailed fetal heart variability analysis from direct electrophysiological recordings correlated to these biomarkers, since they are usually correlated with fetal growth and not the ANS development itself.

The main limitation of our study is the small sample size; however, this is an exploratory study. Although the majority of the high-risk population included pregestational diabetic, we did not have enough sample size to compare them with others so we used it as a confounding variable in our analysis. While pregestational diabetes status had little influence on the relationship between fHRV parameters and some fetal development biomarkers, this could again be attributed to our small sample size. We plan to expand the sample size and perform a prospective study that can be controlled through concurrent enrollment of diabetic (gestational and pregestational) and non-diabetic subjects that meet well-defined inclusion/exclusion criteria. Additionally, for the past few years there has been increasing interest in measuring the precursor of BDNF (proBDNF) and the ratio of proBDNF to mature BDNF, particularly as it relates to the development of depression and neurodegenerative disorders [36]. Measuring proBDNF was beyond the scope of this study since we had limited sample volume and in addition the knowledge on relationship between proBDNF and fetal development is currently limited to make any inferences [37, 38].

Further, we had restricted the criteria to include only scheduled caesarean deliveries in this study to minimize any other induced effects on the inflammatory markers that can happen during vaginal delivery. On the other hand, during the intrapartum period several factors including chorioamnionitis, non-reassuring fetal heart rate patterns, and arrest of labor can affect fHRV parameters and inflammation. Assuming such effects exists, it is unclear how they will influence the levels of correlation between the fHRV parameters and the biomarkers.

In summary, this study indicates the potential relation between fHRV, the inflammatory markers IL-6 and CRP, and the neurotrophic factor BDNF. In a limited sample size, there appears to be a positive correlation of cord-blood BDNF with a parasympathetic activity reflected by a fHRV parameter, and a negative correlation with sympathovagal balance. Meanwhile, cord-blood IL-6 seems to present an inverse relationship in comparison to cord-blood BDNF with respect to parasympathetic activity and sympathovagal balance. These preliminary findings need to be validated in a larger population, but in the future, they could provide a method to track the development of the fetal autonomic nervous system.