Severity of prematurity and age impact early postnatal development of GABA and glutamate systems

Abstract

Gamma-aminobutyric acid (GABA) and glutamatergic system perturbations following premature birth may explain neurodevelopmental deficits in the absence of structural brain injury. Using GABA-edited spectroscopy (MEscher-GArwood Point Resolved Spectroscopy [MEGA-PRESS] on 3 T MRI), we have described in-vivo brain GABA+ (+macromolecules) and Glx (glutamate + glutamine) concentrations in term-born infants. We report previously unavailable comparative data on in-vivo GABA+ and Glx concentrations in the cerebellum, the right basal ganglia, and the right frontal lobe of preterm-born infants without structural brain injury. Seventy-five preterm-born (gestational age 27.8 ± 2.9 weeks) and 48 term-born (39.6 ± 0.9 weeks) infants yielded reliable MEGA-PRESS spectra acquired at post-menstrual age (PMA) of 40.2 ± 2.3 and 43.0 ± 2 weeks, respectively. GABA+ (median 2.44 institutional units [i.u.]) concentrations were highest in the cerebellum and Glx higher in the cerebellum (5.73 i.u.) and basal ganglia (5.16 i.u.), with lowest concentrations in the frontal lobe. Metabolite concentrations correlated positively with advancing PMA and postnatal age at MRI (Spearman’s rho 0.2–0.6). Basal ganglia Glx and NAA, and frontal GABA+ and NAA concentrations were lower in preterm compared with term infants. Moderate preterm infants had lower metabolite concentrations than term and extreme preterm infants. Our findings emphasize the impact of premature extra-uterine stimuli on GABA–glutamate system development and may serve as early biomarkers of neurodevelopmental deficits.

Introduction

Each year in the United States, over half a million children are born preterm, and 1 in every 3 survivors develops debilitating cognitive and behavioral deficits (Laptook et al. 2005; Adams-Chapman et al. 2018; Montagna et al. 2020). Despite improved neonatal intensive care and reduced incidence of severe brain lesions, survivors of preterm birth have a 3–5-fold higher risk for cognitive–behavioral deficits (Laptook et al. 2005; Hintz et al. 2015; Ream and Lehwald 2018). Novel in-vivo tools are needed to detect early perturbations at the level of neurotransmitters that underlie disrupted networks and electro-cortical maturation during premature ex-utero brain development and manifest later as cognitive and behavioral deficits in surviving preterm children.

The third trimester of gestation is a period of rapid development of the principal neurotransmitter systems, including gamma-aminobutyric acid (GABA) and glutamate, and its disruptions after preterm birth have been reported in animal and ex-vivo human studies (Robinson et al. 2006; Shaw et al. 2018; Basu et al. 2021). Perturbations in GABA and glutamate concentrations in key brain regions have been observed in children and adults with cognitive and behavioral deficits (Schur et al. 2016; Dwyer et al. 2018; Godfrey et al. 2018). Because GABA and glutamate signals are obscured by overlapping metabolites on standard magnetic resonance spectroscopy (¹H-MRS), in-vivo human studies need specialized GABA-editing sequences for their reliable quantification (Kwon et al. 2014; Basu et al. 2021). However, due to technical challenges in acquiring GABA-editing ¹H-MRS in infants (Basu et al. 2021), such studies are limited, leaving critical gaps in our understanding of the early postnatal neurometabolic milieu and its perturbations in preterm infants.

J-editing MEscher-GArwood Point Resolved Spectroscopy (MEGA-PRESS) allows simultaneous measurement of in-vivo GABA+ (with coedited macromolecules) and glutamate (Glx; with coedited glutamine) concentrations. We have recently been successful in acquiring reliable data during early infancy and reported normative data on GABA+ and Glx concentrations in 3 key regions of interest (ROI)s, namely, the cerebellum, the right basal ganglia, and the right frontal lobe in healthy term infants (Basu et al. 2022). In term infants, we observed higher GABA+ and Glx concentrations in the cerebellum and basal ganglia (Basu et al. 2022); hubs of regulatory–modulatory centers in the developing brain and functionally more mature at this age in comparison with the frontal lobe (Volpe and Volpe 2018).

With this crucial technical ability, we measured in-vivo GABA+ and Glx concentrations in young infants and investigated for regional differences amongst 3 key regions of the developing preterm brain: the cerebellum, the right basal ganglia, and the right frontal lobe. These 3 ROIs were chosen to represent 3 maturationally and functionally different regions of the developing brain. The cerebellum contains regulatory–inhibitory GABA+ and Glx-rich neurons in the developing brain, matures rapidly during late gestation, and cerebellar injury is independently associated with neurodevelopmental impairment in preterm infants. The basal ganglia are involved in the sensorimotor organization and include an array of nuclei with glutamatergic as well as GABAergic neurons and are vulnerable to hypoxic–ischemic injury associated with premature birth. The frontal lobe is responsible for cognitive and executive function but is relatively less mature in newborns. Herein we sought to determine the relationship of metabolite concentrations with age and prematurity to elucidate the evolving neurometabolic profile of the developing brain during the ex-utero 3rd trimester.

Methods

Participants

Preterm infants born at < 36 completed weeks of gestational age (GA) and admitted in our neonatal intensive care unit (NICU) were prospectively enrolled between 2016 and 2020 in an observational neuroimaging cohort study. Healthy term infants (>38 weeks GA at birth) were recruited from our affiliated obstetric and newborn clinics (Basu et al. 2022). Infants with known genetic syndromes, central nervous system anomalies, or moderate–severe structural brain injury on neuroimaging (identified by >7 score using Kidokoro system scored by an experienced neuroradiologist) (Kidokoro et al. 2013), were excluded from this analysis. Clinical data were collected through medical record review and parental questionnaires. All studies were approved by our Institutional Review Board and conducted per relevant guidelines and regulations. Written informed consent was obtained from the parent(s) of each study participant.

MRI and ¹H-MRS acquisition

Enrolled infants underwent an MRI after reaching term-equivalent age (TEA) during their NICU stay or after discharge as an out-patient visit. MRIs were acquired during their natural sleep using a feed-and-swaddle technique on a 3 T GE Discovery MR750 MR scanner (General Electric Medical Systems, Waukesha, WI) using the identical imaging protocol detailed in previous report (Basu et al. 2022). Briefly, anatomical images were acquired using T2-weighted 3D cube (GE: 3D fast spin-echo sequence with slice thickness of 2 mm. ¹H-MRS data were acquired from voxels placed in 3 key ROIs: a 20 × 15 × 15 mm³ voxel over the right frontal lobe including cortical gray matter and subcortical white matter; a 20 × 20 × 20 mm³ voxel centered on the right basal ganglia; and a 25 × 15 × 10 mm³ voxel placed in the middle of the cerebellum (Fig. 1). We applied standard GE linear shimming before ¹H-MRS acquisitions, without placing saturation bands. The MEGA-PRESS sequence was acquired with the following parameters: TE = 68 ms; TR = 2,000 ms; spectral width of 5,000 Hz; 4,096 points; 256 signal averages. Frequency-selective editing pulses (16 ms duration) were placed at 1.9 ppm during “ON” and 7.5 ppm during “OFF” acquisitions, respectively. J-edited difference (DIFF) spectra were generated by subtracting the “OFF” spectrum from the “ON” spectrum (representative DIFF and OFF spectra are shown in Fig. 2) for each acquisition. Eight unsuppressed water averages were acquired for concentration referencing.

Fig. 1.

Voxel locations of ROIs: a) cerebellum, b) right basal ganglia, and c) right frontal lobe.

Fig. 2.

Representative ¹H-MRS PRESS a) OFF and b) DIFF spectra. Representative spectra with identifiable signal peaks of GABA+ (measured from the DIFF) and Glx, NAA, Cr, and Cho (measured from OFF). The spectra were acquired on a 3 T scanner with TE = 68 ms, TR = 2,000 ms, 256 signal averages, and the editing pulses at 1.9 and 7.5 ppm.

¹H-MRS data pre- and post-processing

The pre- and post-processing methods detailed previously (Basu et al. 2020), include frequency and phase correction (Evangelou et al. 2015) and using the “OFF” and “ON” MEGA-PRESS spectra to generate the DIFF spectrum. Each OFF and DIFF spectra were analyzed using LCModel (Provencher 2001) to measure metabolite concentrations, using the unsuppressed water signal as an internal reference and the same basis-sets used previously (http://purcell.healthsciences.purdue.edu/mrslab/basis_sets.html). The LCModel outputs were visually inspected for artifacts and only those with full-width-at-half-maximum (FWHM) of ≤15 Hz and signal-to-noise-ratio (SNR) ≥ 3 for the OFF spectra and SNR ≥ 2 for the DIFF spectra were included for further analysis.

GABA measurements from MEGA-PRESS DIFF spectra include a contribution from coedited resonances from unspecified macromolecule resonances at 3 ppm and hence are represented as GABA+ (Mullins et al. 2014). Concentrations of Glx (glutamate with contribution from glutamine), NAA (N-Acetylaspartate + N-acetyl-aspartyl-glutamate), Cho (glycerophosphorylcholine + phosphocholine), and Cr (creatine + phosphocreatine) were measured from the OFF spectra LCModel output. All metabolite concentrations are reported in institutional units (i.u.). We a-priori accepted Cramer-Rao lower bound (CRLB) confidence intervals up to 50% for NAA, Cho, and Cr; and up to 100% for GABA+ and Glx due to their lower concentrations; to maximize inclusion of data from spectra otherwise meeting quality parameters, consistent with previous publications (Kreis 2016; Tanifuji et al. 2017; Wilson et al. 2019; Basu et al. 2020; Basu et al. 2022).

Gannet software package (http://www.gabamrs.com/downloads) with modifications pertinent to neonatal MRI was implemented to measure the voxel tissue fractions of cerebrospinal fluid (CSF), gray matter (GM), and white matter (WM) by co-registering voxel masks with tissue segmented 3D T₂ CUBE images (Draw-EM automatic segmentation [biomedia.doc.ic.ac.uk/software/draw-em/] and manual correction in ITK-SNAP [http://www.itksnap.org/pmwiki/pmwiki.php]) (Edden et al. 2014). Cerebellar voxels were segmented into CSF and total brain tissue (GM and WM) fractions due to the intricate folial architecture limiting the accuracy of gray vs. white matter distinction. CSF-corrected metabolite concentrations were determined using the following equation (Maria et al. 2021; Basu et al. 2022):

Statistical analysis

Descriptive statistics were used to summarize the baseline characteristics and ¹H-MRS metabolite measurements. Differences in metabolite concentrations across brain regions and within groups of infants were investigated using ANOVA on Ranks, and pair-wise comparisons performed using Dunn’s method correcting for multiple testing. Correlations between metabolites and age measurements were evaluated using Spearman’s rank correlations. Robust linear regression (M-estimator) was used to assess the trends in metabolite concentrations with postnatal weeks of life (WOL) at MRI and GA at birth for preterm and term neonates to minimize the impact of outlier observations. Sensitivity of the significance in robust regressions was evaluated by median regression with the same covariate setup. Analyses were conducted using R(4.12), and statistical significance was considered for 2-tailed P-values ≤ 0.05.

Results

Descriptive characteristics of the cohort

In this report, we included 123 infants; 48 term (median GA at birth 39.6 weeks) and 75 preterm (median GA at birth 28 weeks) infants with reliable quality of ¹H-MRS acquired at a median 43 vs. 40 weeks of post-menstrual age (PMA) at MRI (P < 0.001) (Table 1). Preterm-born infants were further stratified by GA at birth as moderate (between 28 and 36 weeks) and extreme (≤28 weeks GA) preterm infants. Moderate and extreme preterm infants were similar in sex distribution and PMA at MRI.

Table 1.

Baseline clinical characteristics of term and preterm infants stratified by severity of prematurity.

Clinical parameters Median (IQR)	Healthy term (>38 weeks GA) A (n = 48)	Preterm infants without structural brain injury (<36 weeks GA)			Group comparison P-valuea
		All preterm (n = 75)	Moderate preterm (28–36 weeks GA) B (n = 36)	Extreme preterm (<28 weeks GA) C (n = 39)
Birth weight (in grams)	3,450 (3,180, 3,630)	1,023 (713, 3,170)	1,310 (1,170, 1,470)	723 (613, 970)	<0.001
Gestational age at birth (in weeks)	39.6 (39.0, 40.3)	28.0 (25.3, 30.3)	30.4 (29.0, 31.3)	25.7 (24.0, 26.9)	<0.001
Postnatal weeks of life at MRI (in weeks)	3.6 (1.9, 4.4)	11.9 (9.3, 15.4)	9.2 (7.8, 11.3)	15.1 (12.4, 16.1)	<0.001
Post-menstrual age at MRI (in weeks)	43.3 (41.8, 44.2)	40.0 (38.1, 41.7)	40.0 (38.0, 41.6)	39.9 (38.4, 41.9)	<0.001b
Female sex (%)	21 (44%)	42 (56%)	20 (56%)	22 (56%)	0.07b

IQR: interquartile range, GA: gestational age.

^a P-value for ANOVA between groups of term, moderate preterm, and extreme preterm infants.

^bNot significant for pair-wise comparison between moderate and extreme preterm infants.

Regional CSF-corrected metabolite concentrations

All ROI voxels comprised <10% of CSF on average; CSF-corrected regional metabolite concentrations are reported for preterm and term infants in Table 2 (spectral acquisition parameters in Supplementary Table 1). We observed the highest GABA+ (median 2.44 i.u.), Cho (3.22 i.u.) concentrations, and GABA+/Glx ratio (0.45) in the cerebellum, and the highest NAA concentration (6.16 i.u.) in the right basal ganglia of preterm infants. Glx and Cr concentrations were similar in the cerebellum and right basal ganglia, with the lowest concentrations in the right frontal lobe (Glx 4.02 and Cr 3.25 i.u.). Metabolites demonstrated similar regional profiles in term and preterm infants; with the lowest metabolite concentrations in their right frontal lobes (Table 2). Male preterm infants had higher median GABA+ concentrations in their right frontal lobe compared with females [1.54 (1.2, 1.9) vs. 1.16 (1.0, 1.6)], similar to term infants. Using CRLB cut-off of ≤20%SD, we observed similar (to 50% SD cut-off) median concentrations for NAA, Cho, and Cr (Table 2) and similar regional profiles. Regional voxel SNR (Supplementary Table 1) did not demonstrate significant correlations with respective GABA+ and Glx concentrations (Spearman’s rho < 0.3, P = NS).

Table 2.

CSF-corrected metabolite concentrations (i.u.) stratified by brain regions.

CSF-corrected metabolite concentrations [median (IQR) in i.u.]	Cerebellum	Right basal ganglia	Right frontal lobe	Comparison between regions (Dunn’s method, P-value)
				Cerebellum vs. right basal ganglia	Cerebellum vs. right frontal lobe	Right frontal lobe vs. right basal ganglia
Preterm infants	N = 66	N = 36	N = 64
GABA+	↑2.44 (2.0, 2.9)	2.0 (1.5, 2.3)	1.42 (1.0, 1.8)	0.03	<0.001	0.013
Glx	5.73 (3.8, 8.0)	5.16 (4.3, 6.1)	↓4.02 (3.2, 5.4)	NS	0.004	0.001
NAA	3.75 (2.9, 4.4)	↑6.16 (5.3, 6.9)	3.22 (2.7, 4.0)	<0.001	NS (0.071)	<0.001
NAA_CRLB20%	3.86 (3.3, 4.7)	↑6.16 (5.3, 6.9)	3.33 (2.82, 4.2)	<0.001	NS	<0.001
Cho	↑3.22 (2.8, 3.7)	2.48 (2.4, 2.7)	1.89 (1.7, 2.1)	<0.001	<0.001	0.001
Cho_CRLB20%	↑3.22 (2.8, 3.7)	2.48 (2.4, 2.7)	1.89 (1.7, 2.1)	<0.001	<0.001	0.001
Cr	6.11 (5.3, 6.9)	6.93 (6.4, 7.7)	↓3.25 (2.9, 3.7)	NS	<0.001	<0.001
Cr_CRLB20%	6.12 (5.4, 7.0)	6.93 (6.4, 7.7)	↓3.29 (2.9, 3.8)	NS	<0.001	<0.001
GABA+/Glx	↑0.45 (0.3, 0.7)	0.39 (0.3, 0.5)	0.35 (0.3, 0.5)	0.011	0.003	NS
GABA+/Cho	0.72 (0.6, 0.9)	0.79 (0.6, 1.0)	0.70 (0.6, 0.9)	NS	NS	NS
Glx/Cho	↓1.63 (1.2, 2.2)	2.08 (1.9, 2.3)	2.10 (1.7, 2.6)	0.024	<0.001	NS
NAA/Cho	1.11 (0.9, 1.3)	↑2.54 (2.2, 2.8)	1.83 (1.6, 2.2)	<0.001	<0.001	0.001
Term infants	N = 35	N = 36	N = 27
GABA+	↑2.47 (2.2, 2.8)	2.12 (1.8, 2.4)	1.72 (1.2, 2.0)	0.02	<0.001	0.001
Glx	5.65 (4.4, 6.8)	5.76 (5.2, 6.3)	↓4.10 (3.3, 5.2)	NS	0.001	0.033
NAA	3.70 (3.1, 4.2)	↑6.89 (6.3, 7.6)	4.38 (3.6, 5.2)	<0.001	NS	<0.001
NAA_CRLB20%	3.72 (3.1, 4.2)	↑6.89 (6.3, 7.6)	4.39 (3.6, 5.3)	<0.001	NS	<0.001
Cho	↑3.08 (2.7, 3.4)	2.35 (2.2, 2.5)	1.95 (1.7, 2.1)	<0.001	<0.001	0.001
Cho_CRLB20%	↑3.08 (2.7, 3.4)	2.35 (2.2, 2.5)	1.95 (1.7, 2.1)	<0.001	<0.001	0.001
Cr	6.49 (5.9, 7.1)	6.83 (6.4, 7.4)	↓3.65 (3.2, 4.2)	NS	<0.001	0.001
Cr_CRLB20%	6.49 (5.8, 7.1)	6.83 (6.4, 7.4)	↓3.65 (3.1, 4.3)	NS	<0.001	<0.001
GABA+/Glx	0.44 (0.4, 0.5)	↓0.35 (0.3, 0.4)	0.42 (0.3, 0.5)	NS	0.008	NS
GABA+/Cho	0.83 (0.7, 0.9)	0.87 (0.7, 1.0)	0.88 (0.6, 1.1)	NS	NS	NS
Glx/Cho	↓1.96 (2.5, 2.2)	2.47 (2.2, 2.8)	2.22 (1.9, 2.8)	<0.001	0.02	0.06
NAA/Cho	1.28 (1.1, 1.4)	↑2.93 (2.8, 3.2)	2.30 (2.1, 2.5)	<0.001	<0.001	0.001

N: number of infants with reliable regional spectra, IQR: interquartile range.

↑ and ↓ represent region with highest and lowest metabolite concentration, respectively.

NAA_CRLB20%, Cho_CRLB20%, Cr_CRLB20%: metabolite concentrations excluding those with >20% CRLB; NAA, Cho, Cr data include those up to 50% CRLB.Bold numbers indicate statistically significant differences.

Comparison of CSF-corrected brain metabolites between term and preterm infants, stratified as moderate and extreme prematurity

Preterm infants had lower right frontal GABA+ (1.42 i.u. vs. 1.72 i.u.; P = 0.048) and NAA (3.22 i.u. vs. 4.38 i.u.; P = 0.002) and right basal ganglia NAA (6.16 i.u. vs. 6.89 i.u.; P = 0.007), compared with healthy term infants. Glx concentrations in the right basal ganglia trended lower in preterm infants (5.16 i.u. vs. 5.76; P = 0.087). Stratified by severity of prematurity, moderate preterm infants had generally lower metabolite concentrations than both term and extreme preterm infants (Table 3). GABA+ concentrations were lower in the frontal lobe of moderate preterm (1.07 i.u.) and extreme preterm (1.5 i.u.) compared with term (1.72 i.u.; P = 0.012) infants. Similarly, right basal ganglia Glx concentrations trended lower in moderate preterm infants (4.87 i.u.) and extreme preterm infants (5.39 i.u.) compared with term infants (5.76 i.u.; P = 0.051). Adjusted for PMA at MRI, right frontal GABA+ and NAA concentrations retained significance for differences between prematurity groups of infants. We performed sensitivity analysis excluding infants with regional signal abnormality component of Kidokoro score > 1 (overall Kidokoro injury score < 7) and observed similar relationship of regional metabolite concentrations between categories of infants by GA at birth (data not shown).

Table 3.

CSF-corrected metabolite concentrations stratified by severity of prematurity.

CSF-corrected metabolite concentrations in i.u.	Healthy term infants (n = 48) A		Preterm without structural injury		Comparison between regions P-value
			Moderately preterm infants (n = 39) B1	Extremely preterm infants (n = 36) B2	Overall P-value*	A vs. B1a	A vs. B2a	B1 vs. B2a	Overall P-value (Adjusted for PMA)b
Cerebellum
GABA+	2.47 (2.2, 2.9)		↓2.25 (1.9, 2.7)	2.74 (2.2, 3.6)	0.054	NS	NS	0.07	0.12
Glx	5.65 (4.4, 6.8)		4.82 (3.5, 8.0)	5.86 (3.9, 8.0)	NS				NS
NAA	3.70 (3.1, 4.2)		3.47 (2.8, 3.9)	4.0 (3.0, 4.8)	0.056			0.05	0.01
Cho	3.08 (2.7, 3.4)		3.0 (2.8, 3.5)	3.46 (3.0, 3.8)	0.1	NS	0.034	0.06	0.038
Cr	6.49 (5.9, 7.1)		5.76 (5.1, 6.5)	6.15 (5.8, 7.2)	0.044	0.025	NS	0.1	0.019
GABA+/Glx	0.44 (0.36, 0.51)		0.36 (0.28, 0.65)	0.50 (0.40, 0.66)	NS				NS
GABA+/Cho	0.83 (0.73, 0.91)		0.75 (0.56, 0.87)	0.71 (0.63, 0.96)	NS				NS
Glx/Cho	1.96 (1.56, 2.25)		1.54 (1.19, 2.25)	1.71 (1.18, 2.13)	NS				NS
NAA/Cho	↑1.28 (1.08, 1.42)		1.05 (0.92, 1.25)	1.15 (0.93, 1.31)	0.046	0.005	0.064	NS	NS
Right basal ganglia
GABA+	2.12 (1.8, 2.4)		2.1 (1.7, 2.3)	1.93 (1.4, 2.5)	NS				NS
Glx	↑5.76 (5.2, 6.3)		4.87 (4.0, 5.7)	5.39 (4.3, 6.9)	0.051	NS	NS	0.09	NS
NAA	↑6.89 (6.3, 7.6)		6.30 (5.3, 6.8)	6.08 (4.9, 6.9)	0.015	0.031	0.067	NS	NS
Cho	↓2.35 (2.2, 2.5)		2.47 (2.4, 2.5)	2.68 (2.2, 2.9)	NS	NS			NS
Cr	6.83 (6.4, 7.4)		6.69 (6.6, 7.3)	7.0 (5.6, 8.0)	NS				NS
GABA+/Glx	0.35 (0.3, 0.42)		0.40 (0.33, 0.45)	0.38 (0.28, 0.48)	NS				NS
GABA+/Cho	0.87 (0.72, 1.02)		0.82 (0.65, 1.0)	0.78 (0.59, 0.96)	NS				NS
Glx/Cho	↑2.47 (2.19, 2.80)		2.05 (1.81, 2.31)	2.14 (1.87, 2.31)	0.017	<0.001	0.003	NS	0.023
NAA/Cho	↑2.93 (2.75, 3.22)		2.50 (2.20, 2.81)	2.55 (2.27, 2.60)	<0.001	<0.001	<0.001	NS	0.07
Right frontal lobe
GABA+	1.72 (1.2, 2.0)		↓1.07 (0.9, 1.6)	1.50 (1.2, 1.9)	0.097	0.01	NS	0.014	0.031
Glx	4.10 (3.1, 5.3)		3.95 (3.1, 4.8)	4.14 (3.4, 5.6)	NS				NS
NAA	↑4.38 (3.6, 5.2)		3.12 (2.7, 3.7)	3.24 (2.6, 4.3)	0.011	0.002	0.02	NS	NS
Cho	1.95 (1.7, 2.1)		1.74 (1.6, 2.0)	1.96 (1.8, 2.2)	0.08	0.1	NS	0.013	0.003
Cr	3.65 (3.2, 4.2)		3.22 (2.9, 3.6)	3.40 (2.9, 4.0)	NS				NS
GABA+/Glx	0.42 (0.3, 0.51)		0.30 (0.22, 0.4)	0.40 (0.30, 0.5)	0.039	0.055	NS	0.058	0.07
GABA+/Cho	0.88 (0.62, 1.07)		↓0.60 (0.52, 0.84)	0.83 (0.65, 0.94)	0.002	0.003	NS	0.01	0.042
Glx/Cho	2.22 (1.86, 2.77)		2.22 (1.64, 2.64)	2.08 (1.72, 2.47)	NS				NS
NAA/Cho	↑2.30 (2.07, 2.55)		1.83 (1.63, 2.21)	1.83 (1.56, 2.14)	<0.001	<0.001	<0.001	NS	0.11

↑ or ↓ indicates the infant group with the highest or lowest metabolite concentration, significantly different from other infant groups.

^*ANOVA on ranks overall P-value; apair-wise comparisons adjusted for multiple comparisons by Dunn’s method.

^bRobust median regression, adjusted for PMA at MRI. Bold numbers indicate statistically significant differences.

Relationship between metabolite concentrations with age at MRI

In preterm infants, PMA at MRI demonstrated a significant positive correlation with GABA+ and NAA concentrations (Spearman’s correlation rho 0.2–0.6; cerebellar GABA+ rho = 0.31 with P = 0.033); and a positive trend for Glx (Table 4). Postnatal age measured as WOL at MRI showed a consistent positive correlation with cerebellar and right frontal lobe metabolite concentrations; and generally remained significant after adjusting for GA at birth using robust regression. For term infants, NAA concentrations in all regions correlated positively with PMA at MRI and WOL at MRI (retained significance adjusted for GA at birth), but GABA+ and Glx did not (Supplementary Table 2).

Table 4.

Relationship of regional metabolite concentrations with age parameters of preterm infants.

Metabolites	Spearman’s correlation rho		Linear regression with PMA at MRI	Robust regression modela
	Rho for PMA at MRI (P-value)	Rho for WOL at MRI (P-value)	β (P-value)	β1 for GA at birth (P-value)	β2 for WOL at MRI (P-value)
Cerebellum
GABA+	0.31 (0.033)	0.37 (0.011)	0.13 (0.023)	0.06 (NS)	0.13 (0.016)
Glx	0.19 (NS)	0.26 (0.039)	0.4 (0.036)	0.03 (NS)	0.21 (NS)
NAA	0.43 (<0.001)	0.38 (0.002)	0.24 (<0.001)	0.15 (0.043)	0.21 (<0.001)
Cr	0.52 (<0.001)	0.44 (<0.001)	0.32 (<0.001)	0.26 (0.002)	0.31 (<0.001)
Cho	0.22 (0.08)	0.37 (0.002)	0.07 (0.02)	0.0 (NS)	0.06 (0.043)
Right basal ganglia
GABA+	0.24 (NS)	0.14 (NS)	0.1 (0.07)	0.09 (NS)	0.1 (0.047)
Glx	0.25 (NS)	0.38 (0.025)	0.35 (0.016)	0.15 (NS)	0.28 (0.013)
NAA	0.41 (0.014)	0.2 (NS)	0.39 (0.019)	0.35 (0.008)	0.36 (0.001)
Cr	0.54 (<0.001)	0.35 (0.037)	0.37 (0.04)	0.47 (0.037)	0.37 (0.038)
Cho	0.13 (NS)	0.26 (NS)	0.08 (NS)	0.1 (NS)	0.09 (NS)
Right frontal lobe
GABA+	0.29 (0.029)	0.42 (0.001)	0.08 (0.039)	0.01 (NS)	0.07 (0.023)
Glx	0.21 (NS)	0.26 (0.045)	0.13 (NS)	0.02 (NS)	0.11 (NS)
NAA	0.58 (<0.001)	0.46 (<0.001)	0.36 (<0.001)	0.32 (<0.001)	0.34 (<0.001)
Cr	0.54 (<0.001)	0.45 (<0.001)	0.24 (<0.001)	0.24 (<0.001)	0.24 (<0.001)
Cho	0.20 (NS)	0.43 (<0.001)	0.06 (0.023)	0.02 (NS)	0.05 (0.05)

^aRobust regression model for predicting metabolite concentrations including independent variables gestational age (GA) at birth and postnatal age in weeks of life (WOL) at MRI. Bold numbers indicate statistically significant differences.

Discussion

We report previously unavailable comparative data on in-vivo GABA+ and Glx concentrations in 3 key brain regions of preterm and term infants without major structural brain injury. We observe differences in GABA+ and Glx concentrations in the 3 regions of preterm infant’s brain, likely corresponding to their relative metabolic and functional maturation; similar to that observed in healthy term infants. The observed prematurity-related alterations in neurotransmitter concentrations in the absence of structural brain injury, suggest the influence of subacute and subtle postnatal exposures to cardiorespiratory instability, infection, nociceptive stress, and administration of analgesics and sedative medications.

Unedited ¹H-MRS has been used to measure brain NAA, Cr, and Cho concentrations and interpreted as biomarkers of neuronal integrity; with lower NAA concentrations associated with poorer neurodevelopmental outcomes in preterm infants (Van Kooij et al. 2012; Akasaka et al. 2016; Brossard-Racine et al. 2017; Basu et al. 2019). Consistent with this, we observe lower NAA concentrations across the 3 regions in moderately preterm infants in our cohort (Brossard-Racine et al. 2017; Basu et al. 2019). Since GABA and glutamate signals are overlapped by Cr, Cho, and NAA signals on unedited spectra, high magnetic strength (3–9 T) or GABA-edited ¹H-MRS (like MEGA-PRESS) have been used in animal or older human subjects for their in-vivo measurement. The ability to measure GABA+ and Glx concentrations in older children and adults has allowed for a more targeted investigation of the neurotransmitter milieu with neural connectivity and their association with neurologic disorders (Ende 2015; Schur et al. 2016; Huang et al. 2017; Horder et al. 2018). Similar investigations are needed for high-risk premature infants.

Decreased GABA synthetic enzymes and mRNA of GABA receptors and suppressed glutamatergic neurogenesis have been demonstrated in preterm animal models (Malik et al. 2013; Shaw et al. 2015; Shaw et al. 2018). The developmental trajectory of brain GABA and glutamate concentrations have been described in animal studies, but human in-vivo data during infancy are currently lacking due to the challenges of acquiring time-consuming MEGA-PRESS sequences in non-sedated infants (Robinson et al. 2006; Xu et al. 2011; Malik et al. 2013; Ramu et al. 2016). Overcoming these challenges, we recently reported normative brain GABA+ and Glx concentrations in healthy-term infants (Basu et al. 2022). Our ability to interrogate in-vivo GABA and glutamate systems in preterm infants using GABA-edited ¹H-MRS represents a major advancement for this high-risk population.

In this report, we noted similar regional profiles of GABA+ and Glx concentrations in preterm and term infants (Basu et al. 2022). We speculate that in the absence of structural injury, maturation of key preterm brain regions follows a pattern similar to that in healthy term infants. The higher GABA+ and GABA+/Glx in the cerebellum are consistent with its regulatory–inhibitory function and relatively advanced maturational state during the 3rd trimester (Volpe and Volpe 2018). The basal ganglia are motor sensory hubs, and their metabolic activation during early infancy is reflected in the higher Glx and NAA concentrations, consistent with 2 recent reports from preterm infants (Tomiyasu et al. 2017; Maria et al. 2021). While the overall lower metabolite concentrations in the frontal lobe voxel may reflect its relative immaturity and lower metabolic activation; it could also be the result of the lowest gray matter fraction of the frontal voxel compared with other regions (Supplementary Table 1).

We also investigated the influence of prematurity on brain development by stratifying preterm infants into moderate and extreme preterm groups. The observed lower GABA+, Glx, and NAA concentrations in all 3 regions of the moderate preterm brain (compared with term infants) emphasizes that premature birth may delay or impair typical 3rd-trimester brain maturation at a subcellular level even in the absence of structural injury on conventional neuroimaging. This is consistent with 1 previous report of lower frontal GABA+ in preterm infants (Kwon et al. 2014), and needs further investigation to ascertain the relationship between GABA+ and long-term cognitive and behavioral outcomes.

Longitudinal in-vivo GABA+ and Glx concentrations have not been previously described stratified by severity of prematurity in infants. We hypothesize that the GABA–glutamate system is prematurely stimulated at an immature stage of brain development, more so for the extreme preterm infants. Interestingly, in extreme preterm infants, the GABA+ and Glx concentrations were not significantly different from the term infants and in fact, trended higher than the moderate preterm infants. Although the underlying mechanism remains unclear, we posit that there may be a differential influence of duration (longer for extreme preterm) and degree of stressful exposures (higher for extreme preterm) on premature and aberrant stimulation of the GABA and glutamate system, as indicated by the higher positive regression coefficients for postnatal WOL at MRI (adjusted for GA at birth, Table 2) in preterm infants. This hypothesis is supported by reports of accelerated resting-state connectivity of sensory-motor regions of the developing brain of preterm infants compared with healthy fetuses (De Asis-Cruz et al. 2020). In fact, the observed postnatal rapid increase in GABA+ and Glx concentrations in preterm infants may represent the stimulated neurotransmitter activity that manifests as accelerated resting-state connectivity and warrants further investigation of emerging neurophysiologic function of the measured neurotransmitter concentrations.

Our study’s strength is the acquisition of non-sedated MEGA-PRESS sequences from 3 key regions of the preterm infant’s brain during early postnatal life. Nonetheless, interpretations of our findings should include consideration of several study limitations. The GABA+ and Glx measurements contain a contribution from other coedited molecules and cannot differentiate between synaptic and nonsynaptic metabolites, which limits the interpretation of the actual neurotransmitter pool. Although macromolecule-suppressed GABA estimations have been recently reported in adults, there are limitations and a lack of consensus. For these reasons, we have elected to report GABA+ estimates as such and interpret them for clinical associations (Saleh et al. 2016; Oeltzschner et al. 2018; Duncan et al. 2019; Bell et al. 2021; Craven et al. 2022). Our CSF-correction strategy assumes an equal contribution of gray and white matter towards the metabolite concentration because proportional tissue-specific concentrations of the metabolites remain undetermined for this population. Changes in the tissue water content of gray and white matter of the developing brain may influence metabolite quantification due to its variable T2 relaxation characteristics of the voxel brain tissue. Because the age- and region-specific T2 relaxation times are not established for the newborn/infant brain, we have not performed tissue-specific T2 relaxation correction. However, we have reported GABA+, Glx, and NAA ratios, using Cho as the denominator to partly address the T2 relaxation variability. We prefer Cho as the denominator for metabolite ratios instead of Cr because the latter increases more rapidly with advancing age (β regression coefficients in Table 4) and may interfere with clinical interpretation. We note that relationships of GABA+, Glx, and NAA ratios (with Cho) with brain region and age remain largely in agreement with those observed for the metabolite concentrations. The observed relationship of advancing PMA and postnatal WOL at MRI is cross sectional for this study and may not represent actual longitudinal trajectories after premature birth. Similarly, the influence of postnatal exposures on increasing or decreasing GABA and glutamate concentrations remains unclear and needs further longitudinal investigation. Although the quality parameters of the included spectra meet recent expert consensus statements (Wilson et al. 2019; Choi et al. 2020), about a 5th of the acquired spectra had to be excluded due to motion artifacts and low SNR, challenges inherent for acquisitions from non-sedated infants. Despite these limitations, our findings provide previously unavailable key insights on the underlying neurotransmitter milieu of the preterm brain in the absence of structural injury.

Conclusion

We report in-vivo GABA+ and Glx concentrations from 3 key brain regions of the preterm brain during early infancy reflecting the advanced cerebellar and basal ganglia maturation compared with the frontal lobe, similar to term infants. Even in the absence of structural brain injury, lower metabolite concentrations in moderately preterm infants suggest underlying prematurity-related perturbations that may influence future neurodevelopmental outcomes. The stronger positive relationship of postnatal age at MRI with GABA+ and Glx in preterm infants (but not term infants) supports the hypothesis that postnatal exposures play an important role in functional brain development. Further investigations are needed to determine the prognostic value of early postnatal GABA+ and Glx measures from the preterm brain in identifying infants at risk of cognitive and behavioral deficits.

Authors contribution

Dr Basu conceptualized the design of the study, coordinated, and supervised the data collection, performed extraction of the data and their analyses, interpreted the results, drafted the initial manuscript, reviewed, and approved the final manuscript as submitted.

Dr Pradhan provided technical supervision and innovation with neuroimaging spectral acquisition, post-acquisition processing, conceptualized study design, interpreted the results, reviewed, and approved the final manuscript as submitted.

Dr Sharker was involved in the statistical planning and analyses, critically reviewed the manuscript, and approved the final manuscript as submitted.

Mr Kapse provided technical support for neuroimaging acquisition, post-acquisition processing, conceptualized study design, interpreted the results, reviewed, and approved the final manuscript as submitted.

Ms Catherine Lopez conceptualized the design and coordinated the prospective studies enrolling the healthy newborn subjects, collected data and preliminarily analyzed it, interpreted the results, reviewed, and approved the final manuscript as submitted.

Dr Murnick interpreted the neuroimaging scans, categorized brain injury severity, interpreted the results, critically reviewed the manuscript, and approved the final manuscript as submitted.

Drs Chang, duPlessis, and Andescavage were involved in the conceptualization and design of the study, including overall oversight of study methods, data collection, interpretation of the results, critical review of the manuscript, and final approval of the manuscript as submitted.

Dr Limperopoulos was involved in the conceptualization and design of the study, including overall oversight of technical neuroimaging innovation, procuring data collection, interpretation of the results, critical review of the manuscript, and final approval of the manuscript as submitted.

CRediT taxonomy

Sudeepta Basu (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing—original draft, Writing—review & editing), Subechhya Pradhan (Conceptualization, Data curation, Investigation, Methodology, Resources, Software, Supervision, Validation, Writing—review & editing), Yushuf Sharker (Data curation, Formal analysis, Methodology, Resources, Software, Validation, Writing—review & editing), Kushal Kapse (Data curation, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing—review & editing), Jonathan Murnick (Data curation, Investigation, Methodology, Resources, Supervision, Validation, Writing—review & editing), Taeun Chang (Conceptualization, Investigation, Methodology, Resources, Supervision, Writing—review & editing), Catherine Lopez (Data curation, Investigation, Methodology, Resources, Supervision, Validation, Writing—review & editing), Nickie Andescavage (Conceptualization, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing—review & editing), Adre J. du Plessis (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Writing—review & editing), and Catherine Limperopoulos (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing—review & editing)

Funding

National Institutes of Health—National Heart, Lung, and Blood Institute (award number 1R01HL116585) and Eunice Kennedy Shriver National Institute of Child Health & Human Development (award number R01-HD099393); Intellectual and Developmental Disabilities Research Center (award number 1U54HD090257); National Center for Advancing Translational Sciences (award numbers UL1TR001876 and KL2TR001877); and A. James and Alice B. Clark Foundation award. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Intellectual and Developmental Disabilities Research Center, the National Center for Advancing Translational Sciences, or the National Institutes of Health. Additional support was received from internal departmental funds of Children’s National Hospital, Washington, DC.

Conflict of interests statement

The authors have no financial or non-financial competing interests or relationships relevant to this article to disclose.