Regional brain growth trajectories in fetuses with congenital heart disease

Abstract

Objective:

Congenital heart disease (CHD) is associated with abnormal brain development in utero. We applied innovative fetal MRI techniques to determine whether reduced fetal cerebral substrate delivery impacts the brain globally, or in a region-specific pattern. Our novel design included two control groups–one with and the other without a family history of CHD–to explore the contribution of shared genes and/or fetal environment to brain development.

Methods:

From 2014–2018, we enrolled 179 pregnant women into four groups: “HLHS/TGA” fetuses with hypoplastic left heart syndrome (HLHS) or transposition of the great arteries (TGA), diagnoses with lowest fetal cerebral substrate delivery; “CHD-Other,” with other CHD diagnoses; “CHD-Related,” healthy with a CHD family history; and “Optimal Control,” healthy without a family history. Two MRIs were obtained between 18–40 weeks gestation. Random effect regression models assessed group differences in brain volumes and relationships to hemodynamic variables.

Results:

HLHS/TGA (n=24), CHD-Other (50), and CHD-Related (34) groups each had generally smaller brain volumes than the Optimal Controls (71). Compared with CHD-Related, the HLHS/TGA group had smaller subplate (−13.3% [standard error=4.3%], P<0.01) and intermediate (−13.7% [4.3%], P<0.01) zones, with a similar trend in ventricular zone (−7.1% [1.9%], P=0.07). These volumetric reductions were associated with lower cerebral substrate delivery.

Interpretation:

Fetuses with CHD, especially those with lowest cerebral substrate delivery, show a region-specific pattern of small brain volumes and impaired brain growth before 32 weeks gestation. The brains of fetuses with CHD were more similar to those of CHD-Related than Optimal controls, suggesting genetic or environmental factors also contribute.

Introduction

Congenital heart disease (CHD) is associated with abnormal brain development in utero. Small total brain volume and abnormal cortical folding are detectable by 25–30 weeks gestation.^1–5 Fetal circulatory derangements may directly impair brain development by reducing substrate (e.g., oxygen, glucose, and other nutrients) available to the brain.^{3, 6–9} Two forms of CHD pose a particularly high risk of fetal cerebral substrate deficiency. In d-loop transposition of the great arteries (TGA), the aorta carries substrate-depleted blood from the right ventricle to the brain. In hypoplastic left heart syndrome (HLHS), intracardiac mixing lowers substrate concentration and blood largely reaches the brain retrograde via a hypoplastic aortic arch supplied by the ductus arteriosus. Such abnormal fetal hemodynamics may have a lasting effect on brain structure. In HLHS and related anomalies, smaller neonatal aorta size, a proxy for fetal cerebral blood flow, has been associated with less mature microstructural brain development shortly after birth.¹⁰ Remarkably, this association persists into adolescence.¹¹ Other factors operating in utero may also influence long-term brain structure, including deleterious gene variants,¹² placental dysfunction,¹³ socioeconomic status,^{14, 15} and environmental exposures (e.g., alcohol¹⁶).

Recent advances in fetal neuroimaging have provided new tools to understand fetal brain development in CHD. First, how does abnormal substrate delivery impact brain structure? Does it have a global effect on fetal brain development, or do early injuries and/or deleterious trophic influences impact specific fetal brain compartments causing widespread downstream changes? Animal and human neuropathological data support a specific hypoxic-ischemic vulnerability of neural progenitor cells, premyelinating oligodendrocytes, and subplate neurons, cells which are important for neuronal proliferation, myelination, and cerebral organization.^{6, 17–19} These findings suggest potential cellular targets for fetal neuroprotection (e.g., tetrahydrobiopterin has been proposed to target oligodendrocytes²⁰). Yet existing fetal MRI studies, limited by motion and low signal-to-noise ratio, have not distinguished the transient fetal brain compartments (i.e., the ventricular, intermediate, and subplate zones) where these progenitor/immature cells are abundant. Second, to what extent is abnormal brain structure related to deficient cerebral substrate delivery versus other factors? Published data have variably included control fetuses with a family history of CHD, those being imaged for organ anomalies, and low-risk obstetric clinic volunteers.^{1–3, 21} To distinguish causal factors, control groups should ideally be derived from a comparable genetic and environmental background to the CHD group.

To fill these gaps, we conducted the largest longitudinal fetal brain MRI study of CHD to date. We developed innovative acquisition and postprocessing techniques to define regional brain growth trajectories precisely in isolated CHD including the transient fetal brain compartments of particular interest. We collected echocardiographic data to determine the relationship between fetal cardiac physiology and brain development. We also included two sets of control subjects–one optimal volunteer group and a second with a family history of CHD, the latter expected to have a similar genetic and environmental background to the fetuses with CHD. These data aim to clarify mechanisms of abnormal brain development in CHD, inform strategies for selecting fetuses appropriate for fetal neuroprotection trials, and identify sensitive early measures of therapeutic response.

Methods

Study Design:

From 2014–2018, we performed a prospective, longitudinal cohort study. Subjects underwent up to two fetal MRIs: MRI 1 targeted at 18–30 weeks gestational age (GA) and MRI 2 at 36–40 weeks GA. The MRI 1 timeframe reflects when the subplate and intermediate zones are best distinguished on MRI.^{22, 23}

Participants:

We included four groups: “HLHS/TGA” fetuses with HLHS or TGA, diagnoses with especially low fetal cerebral substrate delivery; “CHD-Other” fetuses with other CHD diagnoses; “CHD-Related Control” healthy fetuses with a family history of CHD; and “Optimal Control” healthy fetuses from pregnant volunteers. All women who underwent clinical fetal echocardiogram at our center were screened for eligibility. Optimal Controls were recruited through hospital intranet advertisements and fliers in clinics.

Inclusion criteria were maternal age 18–45 years and fetal GA 18–30 weeks (determined by the best obstetric estimate²⁴). Exclusion criteria were multiple pregnancy, maternal CHD, mild fetal CHD as defined by Hoffman and colleagues,²⁵ genetic/extracardiac anomaly, brain malformation, congenital infection, MRI contraindication, or clinically inappropriate (e.g., considering termination). Fetuses with a brain malformation or genetic/extracardiac abnormality identified after enrollment were excluded from analyses. Optimal Controls were recruited for other studies in our laboratory but underwent imaging on the same scanners with the same acquisition and postprocessing protocols over the same timeframe as the other groups. They were identified retrospectively and included if they met eligibility criteria, had no incidental anomalies on imaging (to ensure this group reflected normal brain development), and images were of adequate quality.

We recorded fetal sex and maternal race, ethnicity, education, and age at delivery. Genetic testing was obtained through record review and interview. Birth parameters extracted from the clinical record were adjusted for GA. The study was approved by the Boston Children’s Hospital Institutional Review Board, and participants provided written consent.

Fetal Echocardiogram/Doppler:

For the CHD and CHD-Related Controls, structural and hemodynamic data from fetal echocardiogram and Doppler were obtained in accordance with institutional clinical practices and reviewed by a fetal cardiologist (K.G.F.). When multiple echocardiograms were available, we used measures from the echocardiogram proximate to the MRI. Postnatal echocardiographic findings confirmed the diagnosis. Optimal Controls did not undergo echocardiogram.

Cardiac variables included: single ventricle (SV) versus two-ventricle (2V) physiology; presence versus absence of aortic arch hypoplasia; retrograde versus antegrade aortic arch flow; combined ventricular cardiac output (systemic + pulmonary); and expected fetal cerebral blood substrate concentration classified based upon fetal anatomy as normal (substrate-rich), low (substrate-deficient; e.g., TGA), or mixed (e.g., SV). The cerebroplacental ratio calculated from the Doppler data was defined as abnormal if <1.⁷ We also calculated an ordinal “substrate delivery score,” range 0–4, as a single proxy for expected cerebral substrate delivery by summing three factors: aortic arch flow antegrade (0) versus retrograde (1); 2V (0) versus SV (1); and substrate concentration normal (0), mixed (1), or low (2).

MRI Acquisition:

Pregnant women were scanned on a 3-Tesla Siemens MRI scanner, supine if tolerated, otherwise side-lying. Multi-planar repeated T2-weighted half Fourier acquisition single shot turbo spin echo (T2wHASTE) sequences were performed with a 2 or 4 interleaved acquisition; effective echo time 100 and 120 ms; repetition time of 1400–2000 ms; variable field of view based on fetal and maternal size; 2–3 mm slice thickness; no inter-slice gap; 256×204, 256×256, or 320×320 acquisition matrices; and in-plane resolution of 1 mm. Additional sequences were performed to evaluate the rest of the fetal body for other anomalies. Acquisition time was approximately 30 minutes, depending on fetal size and motion.

Image Processing:

T2wHASTE of each subject were processed with our in-house built software and pipeline,²⁶ which involved: super-resolution volume reconstruction with inter-slice motion correction;²⁷ signal intensity non-uniformity correction; registration to our fetal brain MRI atlas common coordinate space; automatic atlas-based segmentation; and manual refinement.²⁶ Under the supervision of a fetal neuroanatomist (L.V.), we updated our published spatiotemporal atlas by segmenting the ganglionic eminence and ventricular zone across ages and measuring the subplate and intermediate zones up to 31.5 weeks GA, as these regions were of interest and able to be visualized. The subventricular zone cannot be accurately delineated on in vivo fetal brain MRI; inner subventricular zone was likely within the ventricular zone label, and outer subventricular zone likely within the intermediate zone label.²⁸ We used the spatiotemporal atlas and individual subject atlases to perform multi-atlas label fusion to generate labels for each subject then manually refined the results. Each subject was segmented using 3–11 atlas images within 1-week of the subject’s GA (e.g., a 27-week subject was automatically segmented using the images from 26-, 27-, and 28-week spatiotemporal atlas templates and individual subject atlases within that age range). Small and narrow structures, such as the hippocampal commissure and fornix were difficult to visualize, thus not refined after automatic segmentation, and measures should be interpreted cautiously. In order to assess reliability of the automatic segmentations for small, subcortical structures, we manually segmented hippocampus, amygdala, and caudate in 10 randomly chosen subjects of varying GA and reconstruction quality. We found strong positive Pearson correlations between the semi-automatic and manual segmentations: hippocampus (r = 0.99, P < 0.001), amygdala (r = 0.84, P = 0.002), and caudate (r = 0.89, P < 0.001).

Regions of Interest:

We analyzed aggregate and individual regions of brain including fetal cortex comprised of cortical plate and hippocampus; developing white matter comprised of subplate zone, intermediate zone, corpus callosum, internal capsule, hippocampal commissure, and fornix; proliferative compartments comprised of ventricular zone and ganglionic eminence; subcortical gray matter comprised of caudate nucleus, lentiform nucleus, and amygdala; diencephalon comprised of thalamus and subthalamic nucleus; brainstem; and cerebellum (Fig 1). Total brain volume included all parenchymal brain regions. Cerebrospinal fluid was comprised of lateral ventricle and other cerebrospinal fluid (third and fourth ventricles and extra-axial fluid). Intracranial volume included total brain and cerebrospinal fluid volume.

Figure 1.

Axial (A), sagittal (B), coronal (C), and 3D (D) images of fetal brain at 26 weeks gestation with regional segmentation as below. Red=cortical plate; Pink=hippocampus; Ocean blue=subplate zone; Yellow=intermediate zone; Orange=corpus callosum; Turquoise=internal capsule; Dark green=ventricular zone; Light green=ganglionic eminence; Purple=caudate nucleus; Gray=lentiform nucleus; Dark red=amygdala; Dark blue=thalamus; Brown=hippocampal commissure; Tan=brainstem; Light pink=cerebellum.

Statistical Analysis:

SAS 9.4 (SAS Institute) was used for all analyses, with the significance threshold set at P<0.05. We compared demographic and medical variables between groups using Fisher’s exact tests or analysis of variance. Comparisons of subjects who underwent two versus one MRI scan were performed using Fisher’s exact tests or t-tests.

Brain volumes.

We examined group differences in brain volumes longitudinally across the GA range using random effects regression models with a random intercept and adjusting for sex and linear and quadratic GA. We performed the primary analyses in two ways: Optimal Control as the reference against which to compare the HLHS/TGA, CHD-Other, and CHD-Related Control groups; and CHD-Related Control as the reference against which to compare the HLHS/TGA and CHD-Other groups. For ease of interpretation, we report results in terms of percent difference in estimated group volumes between an affected 32-week GA male and a 32-week male in the reference group. We selected 32 weeks for the estimate as the midpoint for longitudinal data. Standard error estimates were determined using the bootstrap based on 10,000 replications. Analyses of brain volumes by group included control of false discovery rates to account for multiple comparisons, except as otherwise noted.

Current in vivo MRI cannot distinguish the subplate and intermediate zones after 31.5 weeks GA, thus for longitudinal analyses, the subplate and intermediate zones were merged into a single region, “subplate/intermediate zones”. To examine these two regions separately, we applied linear regression models to measures obtained before 31.5 weeks GA, again adjusting for sex and linear and quadratic GA. For these volumes, we report the percent difference in estimated group volumes between an affected 26-week GA male relative to a same age reference male, with bootstrapped standard error estimates.

Finally, for all of the comparisons described above, we considered group differences in the regional measures independently for right and left hemispheres, and adjustment for total brain volume. We also considered interactions between group and GA to assess potential effect modification on brain growth.

Echocardiogram/Doppler.

We explored associations of brain volumes with echocardiogram/Doppler variables in the fetuses with CHD using random effects regression models with a random intercept and adjusting for sex and linear and quadratic GA. Analyses were restricted to total brain volume and individual brain regions found to have group differences in the primary analysis. Because of the hypothesis-generating nature of these analyses, we did not correct for multiple comparisons. To optimize power while isolating the cardiac variable of interest, analyses of SV/2V status, arch hypoplasia, and retrograde/antegrade flow included those CHD fetuses with normal or mixed (not low) substrate concentration; the analysis of substrate concentration included those CHD fetuses with antegrade (but not retrograde) arch flow. All CHD fetuses were included in analyses of other echocardiogram/Doppler parameters and the calculated substrate delivery score.

Environment/Genetics.

We explored the associations between aggregate brain volumes and maternal education among all subjects using linear regression models with a random intercept and adjusting for group, sex, and linear and quadratic GA. Among the CHD-Related and Optimal Control subjects, we applied linear regression models to assess trends between aggregate brain volumes and the degree of relatedness of the healthy fetus to a family member affected by CHD (sibling/father (0); aunt/uncle (1); cousin/remote (2)), with a random intercept and adjusting for sex and linear and quadratic GA. Given the exploratory nature of these analyses, we did not apply false discovery rate correction but did restrict to aggregate regions to limit multiple comparisons.

Results

Fetal and Maternal Characteristics:

Among 198 eligible women referred for fetal echocardiogram due to suspected fetal CHD, 24 HLHS/TGA and 60 CHD-Other subjects enrolled (Fig 2). After accounting for poor image quality, unable to obtain second MRI, and post-enrollment exclusions, 24 HLHS/TGA subjects contributed 39 MRIs and 50 CHD-Other subjects contributed 77 MRIs to the analysis. Among 206 eligible women whose fetuses had a family history of CHD, 42 CHD-Related Control subjects enrolled with 34 subjects contributing 55 MRIs to the analysis; 71 pregnant women were eligible for the Optimal Control group contributing 91 MRIs. For the 34 CHD-Related fetuses, the affected family member’s relationship to the fetus were sibling (n=20), father (6), aunt/uncle (4), cousin (4), and remote (3). Some fetuses had multiple affected family members.

Figure 2.

Subject screening, enrollment, retention, and imaging quality by group.

Demographic details and qualitative brain MRI findings are in Table 1. Mothers of CHD-Other and CHD-Related Control fetuses were more likely to be of Hispanic ethnicity than those of Optimal Control fetuses. Mothers of Optimal Control fetuses were more likely to have at least a Bachelor’s degree than mothers of any other group. The CHD-Other group was more likely to have undergone genetic testing than the Optimal Control group. The groups were similar with respect to race, maternal age at delivery, and fetal sex, and there were no significant differences in birth parameters. Fetuses with CHD who underwent two MRIs (n=39), compared with those who had one (n=26), did not differ in demographic details or MRI 1 brain volumes. As described above, fetuses with definite brain malformation were excluded. No fetus had parenchymal brain injury. One fetus was excluded from the Optimal Control group due to an incidental finding of mild lateral ventriculomegaly. Echocardiographic and Doppler characteristics for the CHD fetuses are in Table 2.

Table 1.

Maternal and fetal characteristics.

	Optimal Control (n=71)	CHD-Related Control (n=34)	HLHS/TGA (n=24)	CHD-Other (n=50)	Pa
Maternal characteristics
Race					0.78
Black	3 (4)	3 (9)	1 (4)	2 (4)
Caucasian	60 (87)	29 (85)	19 (79)	44 (88)
Other	6 (9)	2 (6)	4 (17)	4 (8)
Hispanic ethnicity	2 (3)	6 (18)	2 (8)	8 (16)	0.03
Education					<0.001
Less than Bachelor’s degree	2 (3)	13 (38)	9 (39)	17 (34)
Bachelor’s degree or higher	67 (97)	21 (62)	14 (61)	33 (66)
Estimated age at delivery, yr	32.5 ± 3.7	31.6 ± 4.3	31.0 ± 5.1	32.7 ± 4.8	0.31
Fetal characteristics
Male sex	44 (62)	19 (56)	18 (75)	29 (58)	0.47
Gestational age, subplate and intermediate zone analyses, wk	25 [22, 27]	27 [26, 28]	28 [25, 30]	27 [25, 30]	0.001b
Gestational age, longitudinal analyses, wk	30 [25, 34]	31 [27, 37]	31 [28, 37]	30 [27, 37]	0.01b
Birth head circumference, cm	34.4 ± 1.0	34.5 ± 1.4	33.8 ± 1.2	33.7 ± 2.3	0.20
Birth head circumference, Z score	0.50 ± 1.04	0.55 ± 1.19	0.16 ± 0.85	0.38 ± 1.15	0.63
Birth weight, kg	3.3 ± 0.3	3.5 ± 0.6	3.2 ± 0.4	3.1 ± 0.7	0.12
Birth weight, Z score	0.45 ± 0.75	0.73 ± 1.00	0.53 ± 0.73	0.45 ± 1.12	0.63
Difference, birth head circumference Z score minus birth weight Z score	0.03 ± 0.97	−0.20 ± 0.75	−0.37 ± 0.65	−0.11 ± 1.04	0.58
Genetic testing	42 (70)	22 (65)	19 (79)	44 (88)	0.047
Qualitative brain MRI findingc	1 (1)	3 (9)	3 (13)	7 (14)	0.03
Diffuse T2 prolongation	0	0	1 (4)	0
Mild lateral ventricle asymmetry	0	3 (9)	0	3 (6)
Mild lateral ventriculomegaly	0	0	0	2 (4)
Possible connatal cyst	1 (1)	0	0	2 (4)
Possible gray matter heterotopia	0	1 (3)	0	0
Possible gyral folding abnormality	0	0	2 (8)	0
Trace hemosiderin in choroid plexus	0	0	1 (4)	0

Values represent n (%) for categorical measures and mean ± SD or median [IQR] for continuous measures.

CHD indicates congenital heart disease; HLHS, hypoplastic left heart syndrome; TGA, d-transposition of the great arteries.

Missing <10% of values except for birth head circumference (Optimal Control, n=12; CHD-Other, n=44), birth weight (Optimal Control, n=13), and genetic testing (Optimal Control, n=11).

^aP values obtained from group comparisons using Fisher’s exact tests or analysis of variance.

^bP values obtained from random effect regression model to account for repeated measures.

^cSome fetuses had more than one finding.

Table 2.

Echocardiographic characteristics for the fetuses with CHD.

	HLHS/TGA (n=24)	CHD-Other (n=50)
Diagnosis
HLHS	13 (54)	0
FAS	0	6 (12)
TOF	0	12 (24)
TGA	11 (46)	0
Other single ventricle	0	13 (26)
Other two ventricle	0	19 (38)
Single ventricle physiology	13 (54)	15 (30)
Substrate delivery
Normal	0	23 (46)
Mixed	13 (54)	23 (46)
Low	11 (46)	4 (8)
Echocardiography characteristics at first MRI
Retrograde aortic flow	13 (54)	8 (16)
Aortic arch hypoplasia	15 (63)	21 (42)
Substrate delivery score
Zero	0	21 (42)
One	0	8 (16)
Two	11 (46)	17 (34)
Three	13 (54)	4 (8)
Abnormal cerebroplacental ratio	5 (23)	8 (17)

Values represent n (%).

CHD indicates congenital heart disease; HLHS, hypoplastic left heart syndrome; TGA, d-transposition of the great arteries; FAS, fetal aortic stenosis; TOF, tetralogy of Fallot; and MRI, magnetic resonance imaging.

Total and Regional Brain Volumes

Longitudinal Analysis.

Using all MRI data across all GAs, we first compared the HLHS/TGA, CHD-Other, and CHD-Related Control groups with the Optimal Control group, then compared the two CHD groups with the CHD-Related Control group.

Compared with the Optimal Control group, the HLHS/TGA group had smaller intracranial volume and total brain volume, but the CHD-Other and CHD-Related Control groups were not significantly different (Table 3). There were no group differences in cerebrospinal fluid volumes.

Table 3.

Estimated group differences for total and regional brain volumes.

	Referent: Optimal Control (n=71)			Referent: CHD-Related Control (n=34)
Volumes (bilateral)	CHD-Related Control (n=34)	HLHS/TGA (n=24)	CHD-Other (n=50)	HLHS/TGA (n=24)	CHD-Other (n=50)
Intracranial volume	−1.7% [1.0]	−6.7% [1.5] **	−1.8% [1.2]	−5.1% [1.5]	−0.1% [1.3]
CSF	−1.5% [1.9]	−5.2% [2.3]	0.5% [1.8]	−3.7% [2.6]	2.1% [2.2]
Lateral ventricle	−4.4% [3.7]	−9.0% [3.6]	−0.7% [3.6]	−4.9% [4.5]	3.9% [4.7]
Other CSF	−1.4% [1.9]	−4.9% [2.4]	0.6% [1.8]	−3.6% [2.7]	2.0% [2.2]
Total brain volume	−1.9% [1.1]	−7.6% [1.4] ***	−3.1% [1.1]	−5.8% [1.5]	−1.3% [1.3]
Fetal cortex	−0.6% [1.5]	−5.4% [1.9] *	−1.7% [1.5]	−4.8% [2.1]	−1.1% [1.7]
Cortical plate	−0.4% [1.6]	−5.4% [2.0] *	−1.5% [1.5]	−5.0% [2.1]	−1.1% [1.7]
Hippocampus	−5.7% [1.2] **	−5.1% [1.5] *	−5.3% [1.1] **	0.7% [1.7]	0.4% [1.3]
All white matter	−2.4% [1.3]	−9.5% [1.5] ***	−4.1% [1.3]	−7.2% [1.7] *	−1.7% [1.5]
Subplate/intermediate zones	−2.2% [1.3]	−9.6% [1.5] ***	−4.1% [1.3]	−7.6% [1.7] *	−2.0% [1.5]
Corpus callosum	−10.1% [1.7] **	−12.6% [1.8] ***	−7.8% [1.8] **	−2.8% [2.0]	2.5% [1.9]
Internal capsule	−1.1% [1.8]	−4.7% [2.1]	−2.0% [1.6]	−3.6% [2.2]	−0.8% [1.7]
Hippocampal commissure	−9.5% [2.6] *	−12.4% [2.4] **	−6.8% [2.7]	−3.2% [2.7]	3.1% [2.7]
Fornix	−4.8% [1.7]	−7.8% [1.8] *	−1.1% [1.8]	−3.1% [2.2]	3.9% [2.2]
Proliferative compartments	−4.2% [1.4]	−10.9% [1.6] ***	−4.6% [1.4]	−7.0% [1.9]	−0.4% [1.8]
Ventricular zone	−3.6% [1.5]	−10.5% [1.7] ***	−4.3% [1.4]	−7.1% [1.9]	−0.7% [1.8]
Ganglionic eminence	−9.6% [2.1] **	−14.9% [2.0] ***	−7.5% [1.9] *	−5.9% [2.6]	2.3% [2.6]
Subcortical gray matter	−5.5% [1.4] *	−7.0% [1.6] **	−3.5% [1.3]	−1.6% [1.9]	2.1% [1.6]
Caudate nucleus	−8.1% [1.6] **	−11.9% [1.7] ***	−6.4% [1.4] **	−4.2% [2.0]	1.8% [1.8]
Lentiform nucleus	−4.1% [1.5]	−4.9% [1.7] *	−2.2% [1.4]	−0.8% [2.1]	2.1% [1.7]
Amygdala	−9.1% [2.4] *	−8.0% [2.6] *	−5.9% [2.3]	1.3% [3.3]	3.5% [3.0]
Diencephalon	−2.8% [1.5]	−5.0% [1.7] *	−3.5% [1.2]	−2.3% [2.0]	−0.7% [1.6]
Thalamus	−2.7% [1.5]	−5.0% [1.7] *	−3.5% [1.2]	−2.4% [2.1]	−0.8% [1.6]
Subthalamic nucleus	−7.8% [3.0]	−6.0% [3.3]	−3.6% [3.0]	1.9% [3.6]	4.6% [3.2]
Brain stem	−0.6% [1.1]	−1.5% [1.2]	−0.8% [0.8]	−1.0% [1.4]	−0.3% [1.1]
Cerebellum	−1.4% [1.7]	−3.1% [2.5]	−1.6% [1.6]	−1.8% [2.7]	−0.2% [1.9]

Values are percent change [bootstrap standard error] for a 32-week gestational age male fetus versus the referent.

^*denotes P<0.05,

^**P<0.01,

^***P<0.001, from false discovery rate-adjusted P values comparing group difference β-estimates.

CHD indicates congenital heart disease; HLHS, hypoplastic left heart syndrome; TGA, transposition of the great arteries; and CSF, cerebrospinal fluid.

Group comparisons were obtained from linear regression models including all MRI data across gestational ages, allowing for a random intercept and adjusting for sex and linear and quadratic gestational age.

For regional brain volumes, compared to the Optimal Control group, the HLHS/TGA, CHD-Other, and CHD-Related Control groups all had smaller hippocampus, corpus callosum, ganglionic eminence, and caudate nucleus volumes. The HLHS/TGA and CHD-Related Control groups in addition had smaller hippocampal commissure and aggregate subcortical gray including specifically the amygdala. Several additional regional volumes were smaller only in the HLHS/TGA group: aggregate fetal cortex and specifically the cortical plate; aggregate white matter and specifically the subplate/intermediate zones and fornix; aggregate proliferative compartments and specifically the ventricular zone; the lentiform nucleus; and aggregate diencephalon and specifically the thalamus (Table 3; Fig 3).

Figure 3.

Estimated group differences in regional brain volumes for the HLHS/TGA group compared to the Optimal Control (top) and CHD-Related Control (bottom) reference groups depicted in axial (left), coronal (middle), and sagittal (right) planes 32-week gestational age fetal MRI. Blue color indicates a relative reduction in brain volume of the structure compared to the reference group; red color indicates a relative increase. Darker intensity reflects a greater magnitude of the estimated group difference in regional brain volume relative to the reference group for a 32-week gestational age male, according to the scale provided. Asterisk denotes significance: * P < 0.05, ** P < 0.01, *** P < 0.001, from false discovery rate-adjusted P values comparing group difference β-estimates (β-estimates in Table 3).

When comparing the two CHD groups to the CHD-Related Control group, the HLHS/TGA group had smaller aggregate white matter volumes and specifically smaller subplate/intermediate zone. The effect size in the proliferative compartments and specifically ventricular zone was comparable to the white matter but not significant (both P=0.07; Table 3).

Exploring group differences separately for the right and left hemispheres, there was no consistent pattern of hemispheric differences distinct from the bilateral analyses (data not shown).

Subplate and Intermediate Zones Analysis.

At 18–31.5 weeks, both the subplate and intermediate zones were smaller in the HLHS/TGA group compared with the Optimal Control and the CHD-Related Control groups. In particular, compared with the CHD-Related fetuses, the HLHS/TGA group had smaller subplate (−13.3% [standard error =4.3%], P<0.01) and intermediate (−13.7% [4.3%], P<0.01) zones, with a similar trend in ventricular zone (−7.1% [1.9%], P=0.07). Volumes of the CHD-Related Control and the CHD-Other groups did not differ from the Optimal Controls (Table 4; Fig 4).

Table 4.

Estimated group differences for subplate and intermediate zones volumes.

	Referent: Optimal Control (n=49)			Referent: CHD-Related Control (n=28)
Volumes (bilateral)	CHD-Related Control (n=28)	HLHS/TGA (n=20)	CHD-Other (n=45)	HLHS/TGA (n=20)	CHD-Other (n=45)
Subplate zone	−0.4% [3.1]	−13.7% [4.1] ***	−4.3% [2.8]	−13.3% [4.3] **	−3.9% [3.3]
TBV-adjusted	0.9 [0.5]	−0.2 [0.6]	0.4 [0.4]	−1.2 [0.6] *	−0.5 [0.5]
Intermediate zone	0.9% [3.3]	−13.0% [4.2] ***	−3.4% [2.7]	−13.7% [4.3] **	−4.2% [3.4]
TBV-adjusted	0.9 [0.5]	−0.3 [0.6]	0.4 [0.4]	−1.1 [0.6]	−0.5 [0.5]

Zone volume values are percent change [bootstrap standard error] for a 26-week gestational age male fetus versus the referent. TBV-adjusted volume values are β-estimate [regression standard error] comparing group differences and adjusting for TBV.

^*denotes P<0.05,

^**P<0.01,

^***P<0.001, from false discovery rate-adjusted P values comparing group difference β-estimates.

CHD indicates congenital heart disease; HLHS, hypoplastic left heart syndrome; TGA, transposition of the great arteries; and TBV, total brain volume.

Group comparisons were obtained from linear regression models limited to 18–31.5 weeks gestation adjusting for sex and linear and quadratic gestational age.

Figure 4.

Unadjusted total and selected regional brain volumes by group as follows: HLHS/TGA (blue closed circle), CHD-Other (blue open circle), CHD-Related Control (red open triangle), and Optimal Control (red closed triangle). Top row depicts total brain and ventricular zone volumes across all gestational ages; bottom row depicts subplate and intermediate zones volumes limited to 18–31.5 weeks gestation. Fitted curves are based on quadratic regression models; non-parallel curves presented only in cases where the interaction between group and gestational age was significant. P values for the associated regression models are included in Tables 3 and 4.

Regional Brain Volumes Normalized by Total Brain Volume

Longitudinal Analysis.

Adjusting the regional measures for total brain volume, we again compared the HLHS/TGA, CHD-Other, and CHD-Related Controls to the Optimal Control group, then each of the two CHD groups to the CHD-Related Control group. In the HLHS/TGA group, only the corpus callosum (β=−0.12 mL [standard error, 0.04]; P=0.02), proliferative compartments (−0.64 mL [0.18]; P=0.01) driven by both ventricular zone (−0.55 mL [0.16]; P=0.01) and ganglionic eminence (−0.09 mL [0.03]; P=0.01), and caudate nucleus (−0.08 mL [0.03]; P=0.01) remained significantly smaller than in the Optimal Control group. In the CHD-Other group, the hippocampus (−0.07 mL [0.02]; P=0.04) and corpus callosum (−0.10 mL [0.03]; P=0.04) remained smaller than in the Optimal Controls, whereas the ganglionic eminence and caudate nucleus no longer met the significance threshold. Comparing the CHD-Related Control to the Optimal Control group, all regions that were smaller in the unadjusted analysis remained smaller after adjusting for total brain volume. In the three-way comparison with the CHD-Related Control group as the reference, no group differences were statistically significant after adjusting for total brain volume.

Subplate and Intermediate Zones Analysis.

After adjusting subplate and intermediate zone volumes measured at 18–31.5 weeks for total brain volume, the HLHS/TGA, CHD-Other, and CHD-Related Control groups did not differ from the Optimal Control group. Comparing the two CHD groups to the CHD-Related Control group, the subplate zone was smaller in the HLHS/TGA group (Table 4).

Brain Growth

Longitudinal Analysis.

There were no differences in the longitudinal total and regional brain growth rates when comparing the CHD-Other and CHD-Related Control groups to the Optimal Control group. However, the HLHS/TGA group had slower growth of lateral ventricles (β=−0.106 mL/week [0.036]; P=0.03), hippocampus (−0.015 mL/week [0.005]; P=0.03), corpus callosum (−0.026 mL/week [0.008]; P=0.01), ganglionic eminence (−0.015 mL/week [0.005]; P=0.01), and caudate nucleus (−0.019 mL/week [0.005]; P=0.01) than the Optimal Control group There were no group differences in total or regional brain growth rates when comparing either of the two CHD groups to the CHD-Related Control group.

Subplate and Intermediate Zones Analysis.

At 18–31.5 weeks GA, the slopes of both the subplate and intermediate zone volumes suggested slower regional brain growth for the HLHS/TGA group compared with both the Optimal Control (subplate zone: β=−1.02 mL/week [0.33], P<0.01; intermediate zone: −0.96 mL/week [0.25]; P<0.001) and the CHD-Related Control groups (subplate zone: −1.19 mL/week [0.46]; P<0.05; intermediate zone: −0.98 mL/week [0.36]; P<0.01). Slopes of the CHD-Other group did not differ from either control group.

Cardiac Factors

Brain Volumes.

In longitudinal models including all MRI data across GAs, smaller total brain volume was associated with SV status (β=−10.0 mL [4.7]; P=0.04); smaller subplate/intermediate zone volume was associated with SV status (−6.6 mL [2.6]; P=0.01) and lower substrate delivery score (−2.2 mL per point [1.1]; P=0.046); smaller ventricular zone volume was associated with lower cardiac output (−1.8 mL [0.6]; P=0.006); and smaller corpus callosum and hippocampal commissure volumes were associated with aortic arch hypoplasia (corpus callosum: −0.10 mL [0.05]; P<0.05; hippocampal commissure: −0.03 mL [0.01]; P=0.01). At 18–31.5 weeks GA, when subplate and intermediate zones can be distinguished, neither subplate nor intermediate zone volume was associated with echocardiographic or Doppler variables.

Genetic/Environmental Factors

Higher levels of maternal education were associated with larger subcortical gray matter (college β=0.35 mL [0.14]; P=0.014; graduate degree 0.46 mL [0.14]; P<0.001). The pattern of group differences was not appreciably changed by including maternal education in longitudinal models predicting total and regional brain volumes or in cross-sectional models predicting subplate and intermediate zone volumes. A more remote family history of CHD was associated with larger brain volume of the subcortical gray matter (i.e., cousin/remote > aunt/uncle > sibling/father; 0.15 mL per point [0.05]; P=0.006).

Discussion

Our longitudinal, fetal brain MRI study identified regional vulnerability in the subplate, intermediate, and ventricular zones, regions that contain subplate neurons, premyelinating oligodendrocytes, and neural progenitor cells thought to be vulnerable to hypoxia-ischemia. Fetuses with HLHS/TGA had larger volume reductions than fetuses with other forms of CHD in comparison to both control groups. Among fetuses with CHD, reduced volumes of these regions were associated with echocardiographic measures related to cardiac output and substrate concentration. Volumetric reductions were attenuated when the comparison group had a family history of CHD, highlighting a potential contribution of genetic and/or environmental factors to fetal brain development. Taken together, these data are consistent with the hypothesis that CHD is associated with a genetic and/or environmental background of developmental brain differences with superimposed effects of oxygen and/or nutrient deficiency on selectively vulnerable fetal brain structures.

Our findings bolster existing evidence that fetal circulatory disturbances directly impact brain development in CHD through lowering fetal cerebral substrate delivery. Cerebral substrate delivery depends on both cerebral blood flow and substrate concentration. Fetal neuroimaging work using T2* and cine phase-contrast MRI suggest lower cerebral oxygen delivery in fetuses with SV heart disease and TGA.^{3, 29} Reduced fetal cerebral oxygen delivery and consumption have been associated with smaller total brain volume, as has been a lower fraction of combined cardiac output through the aortic valve.^{1, 3} In our cohort, smaller total brain and subplate/intermediate zone volumes were associated with SV status and a worse substrate delivery score, measures that reflect lower cerebral blood flow and substrate concentration. Furthermore, the HLHS/TGA group, expected to have particularly low fetal cerebral substrate delivery, had larger magnitude and more diffuse reductions in brain volumes than the CHD-Other group.

We found regional vulnerabilities in the subplate, intermediate, and ventricular zones. Subplate and intermediate zones were smaller and grew more slowly in the HLHS/TGA compared with both control groups, and the ventricular zone showed a comparable, but non-significant, effect size. These results align with previously demonstrated reductions in developing white matter volumes,^1–3 though they contrast with those of Olshaker and colleagues who found preserved cerebral volumes but smaller cerebellar volumes.²¹ This discrepancy may relate to differences in study populations, as Olshaker compared a heterogeneous CHD cohort to a reference group undergoing brain MRI for clinical indications.

Abnormalities in subplate, intermediate, and ventricular zone development could impact long-term brain structure.^{30, 31} The subplate zone is comprised of subplate neurons, a rich extracellular matrix, and “waiting” thalamocortical and cortico-cortical fibers.^{32, 33} The subjacent intermediate zone contains migrating neurons, axons, and premyelinating oligodendrocytes. Early in gestation, the ventricular zone generates neural progenitor cells and glia. Radial glia migrate to form a scaffold within the intermediate and subplate zones, which neurons traverse towards the cortical plate. Subplate neurons send axons to facilitate cerebral cortical organization and thalamocortical/cortical-cortical connectivity. Oligodendrocytes begin myelination in the second trimester. The volumetric reductions in our dataset may reflect disruption to this highly orchestrated pattern and contribute to abnormal connectivity and cortical development detected in children/adolescents with CHD.^{10, 11, 34–37}

Animal models and data from preterm neonates suggest that components of subplate, intermediate, and ventricular zones are vulnerable to hypoxia-ischemia. Hypoxia-ischemia alters the maturational trajectory and synaptogenesis of subplate neurons.^{17, 38–40} Premyelinating oligodendrocytes are exquisitely sensitive to hypoxia-ischemia, which causes maturational arrest and failure to produce myelin.⁴¹ Morton and colleagues developed a perinatal porcine model of fetal cerebral hypoxia-ischemia in CHD and found vulnerability of neural progenitor cells associated with abnormal cortical development.⁶ A recent fetal sheep hypoxia model also found reduced neuronal density, myelination, and abnormal microglial morphology after prolonged fetal cerebral hypoxia.^{9, 42} Fetal brain MRI cannot discern these neural components at the cellular level and lacks resolution to distinguish the subventricular zone from the surrounding intermediate and ventricular zones. However, our findings are consistent with the hypothesis that subplate neurons, premyelinating oligodendrocytes, and neural progenitor cells are selectively vulnerable to substrate-deficiency in fetal CHD.

To our knowledge, we are the first to compare fetal MRI data in CHD with two separate control groups: one with a family history of CHD and one without. Control groups in prior studies have variably included fetuses with a family history of CHD, those from low-risk obstetric clinics, and fetuses being imaged for non-cardiac anomalies.^{1–3, 21} We found widespread differences in brain volumes in the HLHS/TGA compared to the Optimal Controls, but reduced volumes in only select vulnerable regions when compared to the CHD-Related Controls. Further, the brains of fetuses with a family history of CHD were broadly smaller than those of the Optimal Control group. The disparate findings between comparisons to the two control groups suggests that data utilizing heterogeneous control groups should be interpreted cautiously, as the findings may not reflect a specific hemodynamic effect of CHD, but rather additional genetic and/or environmental differences.

Genetics/genomics may contribute to the discrepant findings between control groups. CHD has been associated with a variety of genetic abnormalities, including variants associated with neurodevelopmental disorders.^{12, 43, 44} If the CHD-Related Control cohort is enriched for these same variants as the CHD groups, that genetic background could impact fetal brain development. No published data examine the association between a family history of CHD and fetal brain development. One study reporting small birth head circumference in CHD found that the results did not differ when comparing to a general population versus sibling controls.⁴⁵ This finding suggests either that birth head circumference may not be sensitive to detect subtle differences in brain growth, or that a family history of CHD may not impair global fetal brain growth.

An alternative explanation for the discrepancy between comparisons to the two control groups is confounding by socioeconomic status or other environmental influences. Socioeconomic status predicts neurodevelopmental outcome and brain structure after birth.^{14, 15, 46, 47} In our cohort, the Optimal Control group had higher levels of educational attainment but comparable maternal age to the CHD-Related group. A higher level of education may be associated with other factors that influence fetal health, such as access to prenatal care, stress, or nutrition.^{46, 48–51} It is notable however, that we found few associations between either maternal education or family history proximity and fetal brain volumes. Adjusting for maternal education in the main models did not change study inferences. Future studies with larger cohorts should further investigate the relationship between fetal brain development in CHD and genetics/genomics, socioeconomic status, and environmental factors such as maternal health (e.g., hypertension, diabetes), prescribed and/or illicit substance/alcohol use, and toxin exposure (e.g., plasticizers).

Limitations

Our study has several limitations. First, the sample size did not allow for detailed subgroup analysis. We chose a two-pronged strategy, combining the HLHS and TGA into one group for the primary analysis while exploring specific hemodynamic variables in a secondary analysis. Future larger cohorts are needed to examine the contributions of fetal hemodynamic factors to brain development within specific CHD diagnoses. Second, we did not directly measure fetal cerebral oxygen delivery, instead extrapolating from the echocardiogram anatomy. With present technology, the time to acquire detailed structural imaging and directly measure cerebral oxygenation in a mobile fetus precluded acquiring both in the same session. Thus, we cannot determine whether the causal pathway included reduced cerebral oxygen delivery, reduced nutrient delivery, and/or other factors. Third, fetal brain MRI cannot determine the specific cellular and extracellular components of the transient fetal brain compartments that contributed to volumetric differences. Correlation of our results with future neuropathologic studies could elucidate the neurobiological mechanisms. Fourth, small and narrow structures which are difficult to visualize, such as the hippocampal commissure, were automatically segmented and should be interpreted cautiously. Finally, although we screened carefully to exclude fetuses with suspected genetic conditions, all subjects did not undergo genetic testing. Our sample may include some fetuses with genetic variants that were not clinically detected.

Conclusions

In summary, we demonstrate that abnormalities in fetal brain growth trajectories in CHD vary across brain regions and emerge before 32 weeks GA. Patterns of regional involvement are consistent with the hypothesis that subplate neurons, premyelinating oligodendrocytes, and neural progenitor cells are selectively vulnerable to substrate-deficiency in fetal CHD and may present potential neuroprotective targets. We also showed for the first time that the brains of fetuses with CHD were more similar to those of fetuses with a family history of CHD than to those of an optimal comparison group, suggesting that disrupted fetal brain development demonstrated in prior studies may reflect not only the presence and type of heart disease, but also previously unmeasured confounders in genetics and environment. Overall, our data suggest that widespread differences in fetal brain development may reflect genetic/environmental factors with superimposed regional patterns of vulnerability in forms of CHD with the most severe fetal cerebral substrate deficiency.