Social brain circuitry and social cognition in infants born preterm

Magnetic resonance imaging (MRI) studies show that the most common structural observations in PT infants include diffuse white matter (WM) and gray matter (GM) abnormalities. Periventricular leukomalacia (PVL) is characterized by both focal and diffuse lesions to the brain’s white matter. Inder et al. [89] collected quantitative volumetric MRI in preterm infants (birth GA < 32 weeks) with prior evidence of PVL. At term, these infants displayed a reduction in cerebral cortical (but not subcortical) gray matter, along with a reduction in the volume of total brain myelinated (but not unmyelinated) white matter. PVL was, until recently, the most commonly diagnosed brain injury in PT infants. The recent decrease in the prevalence of cystic PVL is thought to be associated with several factors, including reductions in the duration of mechanical ventilation and the incidence of bacterial sepsis [90]. It has since been recognized, however, that white matter damage often occurs in the absence of the cystic regions of necrosis seen in classic PVL [91, 92]. A later study [93] reported reduced cerebral cortical gray matter volume and deep nuclear gray matter volume in 119 preterm infants (birth GA ≤ 32 weeks, birth weight ≤ 1500 g), as compared to 21 full-term infants, and found that these abnormalities correlated with moderate to severe neurodevelopmental disability at 1 year of age. Brown et al. [94] imaged a cohort of very preterm (VPT, birth GA < 30 weeks or birth weight < 1250 g) infants and found that lower composite neurobehavioral scores at term were related to the degree of white matter abnormality observed via MRI, and correlated most strongly to abnormal WM signal and reduced WM volumes, rather than the grade of GM abnormality. In aggregate, these findings suggest atypical gray and white matter development and associated behavioral abnormalities occur even in PT infants who do not meet criteria for canonical insults such as PVL. Neurodevelopmental research has traditionally been focused on infants with identifiable injury or severe symptomatology, but there is increasing recognition of the need to identify infants who may manifest more subtle impairments.

The brain undergoes rapid and dynamic development during the latter two trimesters. In terms of gross structure, total brain volume increases and ventricle size decreases during the second trimester, and while major fissures begin to emerge, the cerebral surface remains largely smooth until the end of the second trimester [95]. A small amount of cortical folding occurs by 25 weeks, and folding intensifies in the third trimester [96]. This folding occurs in an orderly fashion, with many primary sulci forming during the second trimester, and tertiary sulci appearing during the third trimester and after term [97]. Sulci and fissures emerge in a posterior to anterior fashion: the parieto-occipital, calcerine, and cingulate sulci around 20–22 weeks GA; the central, interparietal, and superior temporal sulci by 25 weeks GA, and the precentral, post-central, superior frontal, and middle temporal sulci around 24–28 weeks [98]. Between 18 and 40 weeks GA, fetal white matter volume shows a 22-fold increase, along with a 21-fold increase in cortical gray matter and a 10-fold increase in deep subcortical structures [99]. Presumably, birth prior to 40 weeks may disturb the typical progression of neural circuit organization. Structural MRI studies have indeed shown abnormalities in specific structures and regions in the preterm brain, including areas associated with the putative social brain. The majority of these studies have imaged former PT infants at school age or later. This later-stage imaging (see Box 1) introduces difficulties in interpretation—while this body of work suggests lifelong consequences for a portion of the PT-born population, the substantial interim between birth and assessment introduces difficulty in specifying the contributions of prematurity, per se, as opposed to potential downstream or tangential effects. For this reason, this portion of the review will focus on data collected during infancy, rather than imaging acquired later in childhood or in adulthood.

As mentioned above, the set of structures associated with social cognition is widely distributed, and emerging evidence suggests that prematurity is associated with altered brain volumes in many of these regions. Thompson et al. [100] compared 202 preterm (birth GA < 30 weeks and/or birth weight < 1250 g) and 36 term infants (scanned at term equivalent age (TEA)) and found that PT infants showed reduced cortical gray matter and unmyelinated WM volumes in orbitofrontal regions. Similarly, Ball et al. [101] collected scans from 71 preterm (birth GA < 136 weeks) infants at TEA and found that degree of prematurity was associated with volume reduction in the orbitofrontal lobe. The orbitofrontal cortex seems to be a key for self-monitoring [102] and emotional self-regulation [103], amongst many other behaviors.

In the parietal region, Ball et al. [101] also found that increasing prematurity was associated with reduced posterior cingulate cortex (PCC) volume. This finding was echoed by Gousias et al. [104], who compared scans from 15 PT infants (<36 weeks GA, imaged at TEA) and 5 FT infants and also found that PT infants had relatively smaller PCC volumes. The PCC is thought to play a crucial role in the so-called “default mode network” (DMN) [105]. This network includes the PCC, medial prefrontal cortex (mPFC), and medial temporal lobe (MTL) and several studies have pointed to its role in social-cognitive processing [106, 107].

Significant volumetric differences have also been observed in the occipital region in the brains of PT infants. Peterson et al. [108] scanned 10 PT (birth GA not specified) and 14 term infants once between 34 and 42 weeks GA and calculated volumes for cortical gray and white matter parcellated into 8 subregions. In this study, PT infants were observed to have reduced GM volumes in the parieto-occipital and inferior occipital cortices. Similarly, Thompson et al. [100] found that PT infants showed reduced deep nuclear gray matter in the parieto-occipital region. The extrastriate body area and occipital face area, which would likely be captured in the “inferior occipital” subdivision, have been implicated in perceiving human bodies and body parts [109] and accurate face perception [110], respectively. The area designated as “parieto-occipital” in these studies also may have been large enough to capture the temporoparietal junction, an area associated with numerous aspects of social cognition including theory of mind [111], empathy [112], and moral judgments [113].

Many of the structures associated with the social brain are concentrated in the temporal lobe, including the amygdala [114], the superior temporal sulcus [115] and superior temporal gyrus [116], the fusiform face area [117], and the temporal pole [116], among others. In Gousias’ study [104], several significant differences were pinpointed in the temporal region. Here, PT infants had relatively smaller volumes in the middle and inferior temporal gyri, as well as the anterior temporal lobe. Interestingly, in this sample, the PT group (as compared to the FT group) showed larger volumes in the amygdala, as well as the left superior temporal gyrus (though it should be noted that none of these group differences remained significant after correcting for multiple hypothesis testing). The amygdala is one region consistently implicated in processing social information, including emotional responses [118], detection of socially salient information [114], and social affiliative behavior [119] (see [120] for a review of amygdala function). The temporal lobe has also been a region of strong interest in follow-up imaging conducted after infancy (see Box 1).

While volumetric findings can admittedly be difficult to interpret, the research reviewed above suggests that preterm birth is associated with atypical cortical and subcortical development. This altered development could be a result of disruptions to the typical developmental processes occurring during the second and third trimesters. These alterations in both gray and white matter appear in several brain regions believed to underlie processing of social information.

Much of the most recent work on preterm brain development has used a method called diffusion weighted imaging (DWI) to investigate microstructural development.

Diffusion weighted imaging (DWI) measures the restricted diffusion of water molecules. The rate of diffusion is affected by tissue structures. For example, water moves more readily along axonal bundles than perpendicular to them and diffuses isotropically in cerebrospinal fluid. The directionality of anisotropic diffusion of any given voxel can be characterized by applying a tensor model (i.e., diffusion tensor imaging or DTI). By providing information about the spatial distribution of water diffusion in a given voxel, diffusion imaging can serve as a tool for quantitatively characterizing brain microstructure. Common diffusion metrics include axial diffusivity, radial diffusivity, and mean diffusivity. Axial diffusivity (AD) refers to diffusion parallel to the principle eigenvector of the three-dimensional voxel, while radial diffusivity (RD) denotes diffusion in the direction orthogonal to the principle eigenvector. Mean diffusivity (MD) or apparent diffusion coefficient (ADC) represents the average of the two fractional anisotropy (FA) reflects the degree of diffusion along the principle axis relative to the other directions. FA is a scalar metric with values ranging from 0 (unrestrained or “isotropic” diffusion) to 1 (completely constrained or “anisotropic” diffusion). Opinions differ somewhat as to which aspects of the brain’s microstructural architecture are reflected best by each measure. It is thought that MD is most closely related to overall water content in the brain [121], while FA represents an index of the brain’s microarchitecture that is influenced by axonal density, axonal diameter, cell membrane and microtubule composition, myelination, and dendritic cytoarchitecture [80, 121‐123].

In terms of gray matter, postmortem studies of fetal brains show an increase in regional cortical FA from 15 to 28 weeks of gestation, followed by a decrease through 36 weeks [124]. This finding is consistent with the migration of neurons from the germinal matrix, as it is thought that the bulk of cortical neurons migrate along radially organized glial fibers early in gestation (Rakic, 1988, 2003; Rakic and Kornack, 2001; as cited in [124]), perhaps contributing to the increased cortical FA observed prior to 27 weeks. At this point, these glial cells retract (Volpe, 2001; Rakic 2003; as cited in [124]), ostensibly resulting in a decline in cortical FA. Thus, gray matter microstructure undergoes significant change during the period of gestation disrupted by preterm birth.

McKinstry et al. [122] attempted to characterize water diffusion in vivo in neonatal cortical gray matter by imaging 24 PT and FT infants (birth GA 26–41 weeks) in the first 36 h of life. These studies found that (similar to Huang et al. [95]) the cerebral cortex shows a pronounced radial organization by 26 weeks of gestation, but this radial organization disappears by term, as evidenced by a decrease in averaged water diffusion coefficient. Ball et al. [123] imaged preterm infants (birth GA 23–35 weeks), 47 of whom were scanned once, and 9 of whom were scanned twice. Between 27 and 46 weeks post-conception, a decrease in cortical FA was observed until week 38. Higher initial FA values and rates of change were seen in gyri, frontal and temporal poles, and parietal cortex, with lower values and rates of change in sulcal, perirolandic, and medial occipital cortex. Kersbergen et al. [125] looked at longitudinal change in FA in 122 distinct regions of the brain in infants between 30 and 40 weeks of postmenstrual age. The sample included infants born prior to 28 weeks of gestation and was limited to individuals without cerebral injury and with a normal neurodevelopmental outcome (as characterized at 15 months of age with the Griffiths Mental Development Scales), who were scanned longitudinally around 30 weeks of gestation and again at term equivalent age. In cortical brain areas, frontal areas showed a slight increase in FA, while temporal-occipital areas exhibited a slight decrease in FA. The majority of brain regions showed a clear decrease in mean, radial, and axial diffusivity, though the degree of decrease was heterogeneous across regions. Similar to the pattern of FA change, increases in MD, AD, and RD were seen in the occipital, temporal, and frontal regions. As described above in the context of structural imaging, the putative social brain encompasses cortical areas spread across the brain. As such, microstructural alterations in these areas (as indexed, in this case, by altered diffusion metrics) might contribute to atypical patterns of social cognitive development.

White matter has been of particular interest in the realm of preterm brain development, for multiple reasons. As described earlier, it appears to be acutely vulnerable to insult. DWI provides especially important information about white matter development during the perinatal period in the early stages of myelination [79]. Studies of postmortem fetal brains suggest that white matter fibers transition from a dominant radial organization around 15 weeks of gestational age to a more laminar and radial architecture through the duration of the second trimester [95], Meanwhile, white matter tracts considered crucial to early social-cognitive behaviors (including the corpus callosum, the uncinate, and the inferior longitudinal fasciculus) undergo rapid change from 13 to 22 weeks of gestation [95]. More generally, imaging from fetal brain specimens and full-term neonates [95, 126] indicates that limbic fibers (including the stria terminalis, fornix, and cingulum) are well formed by 19 weeks of gestational age. Association fibers (including the uncinate fasciculus, inferior fronto-occipital fasciculus, and superior and inferior longitudinal fasciculi) seem to develop relatively more slowly [95]. Based on three-dimensionally reconstructed WM tracts of fetal specimens, the uncinate is visible by 15 weeks of gestation, but the superior longitudinal fasciculus is still not prominent by term [95]. White matter FA appears to increase rapidly from term through the first 2 years of life, at which point it slows but continues to increase through childhood [79, 127].

Critically, several of the tracts developing largely in the second and third trimesters have been consistently implicated in various aspects of social processing. The corpus callosum is visible in rudimentary form by the twelfth week of gestation [128], but undergoes exuberant proliferation during the third trimester followed by a period of axonal pruning until the second month post-term [129]. Much of the work investigating the role of the corpus callosum in social functioning comes from studies of individuals in whom the fibers of the corpus callosum fail to develop, a condition known as agenesis of the corpus callosum (AgCC). Parents of children with AgCC report impairments in social competence, introspection, social judgment and planning, and poor emotional communication [130]. Abnormalities of the corpus callosum have also been associated with the range of social deficits observed in autism, specifically measures of social responsiveness [131]. One study of corpus callosum diffusion at TEA in 58 PT infants (divided into groups of 23–25, 26–29, and 30–33 weeks birth GA, all without apparent WM lesions) found that lower birth GA was associated with lower FA in the posterior corpus callosum, such that FA decreased linearly with birth GA [132]. Another study using DTI in 44 PT (birth weight between 600 and 1250 g) and 41 FT adolescents found that prematurity was associated with lower FA in multiple regions including the splenium of the corpus callosum [133]. Eikenes et al. (2011) also found reduced FA in the corpus callosum of 49 VLBW (birth weight < 1500 g) PT-born 12-year olds, when compared to 59 term-born controls [134]. A recent meta-analysis [135] incorporated 232 original studies, for a total of 513 PT-born children, adolescents, and young adults and 309 healthy controls. The analysis used an activation likelihood estimate (ALE) procedure to identify regions of abnormal FA in the corpus callosum. The ALE analysis yielded 11 regions of decreased white matter FA in the PT versus FT group, including the bilateral splenium of the corpus callosum, the bilateral external capsule, the left superior fronto-occipital fasciculus, the left posterior thalamic radiation, the right superior longitudinal fasciculus, the genu of the corpus callosum, the left cingulum, the left posterior corona radiata, and the left posterior limb of the internal capsule (also seen in [136]). Four regions of increased FA were observed, including the bilateral posterior corona radiata and the anterior corona radiata. The authors interpret these findings to mean that the mechanisms underlying alterations in FA in the preterm brain may vary during different stages of white matter development. The areas of increased FA were smaller than the areas of decreased FA (192 and 4360 voxels, respectively), and the authors note that the findings of decreased FA are more consistent with prior work. The regions of increased FA seem harder to explain; it is suggested that these increases may in fact be due to differences in methodology between studies.

The uncinate is a bidirectional long-range pathway that extends from the anterior temporal lobes and amygdala to the lateral orbitofrontal cortex [137]. While the exact function of the uncinate is under debate [138], it serves to connect areas believed to be vital to various social-cognitive functions, and has been associated with behaviors such as joint attention [81] and sensitivity to social rewards [139]. Abnormalities in the uncinate have been associated with various psychiatric disorders (see [140] for a review), and it has been hypothesized that the slow time course of uncinate development may make it particularly susceptible to perturbation contributing to social-emotional problems. There has been little investigation of the uncinate in PT infants, but in addition to their findings related to the corpus callosum, Eikenes et al. [134] and Mullen et al. [133] found reduced FA in the uncinate of adolescents born preterm.

Recent years have seen a striking increase in work investigating diffusion in the PT brain, and as of yet there is little consensus as to the precise structures or circuits most often affected by prematurity. As Li [135] notes, this may be due at least in part to differences in methodology, including the population of interest, the time at which images are collected, and the manner in which brains are segmented into regions or networks. What is clear is that even uninjured PT-born infants show differences in brain connectivity, both structurally and functionally (as described below).

One burgeoning area of research involves functional imaging in preterm infants. The bulk of this work has been conducted with adolescents or adults born preterm, in which case it is challenging to draw specific conclusions given the amount of time lapsed, but some groups have attempted to characterize functional activity closer to the time of birth. As described in the diffusion imaging section, prematurity has been found to be associated with alterations in the brain’s connectomic architecture. Another way to characterize the development of the brain’s neural circuitry is to examine functional connectivity through methods such as resting state fMRI (functional magnetic resonance imaging). Echoing the aforementioned findings related to structural brain changes in the third trimester, this is also a time of great change in the organization and connections in the brain’s functional networks. In a cohort of 40 preterm and full-term infants (birth GA 25 to 41 weeks) imaged between 31 and 42 weeks PMA, higher PMA was associated with greater functional connectivity [141]. There were also regional differences, such that age-dependent increases in connectivity were greater in primary sensory and motor regions and lesser in default mode and executive-control regions. Doria et al. [142] collected resting state fMRI (rs-fMRI) data in 70 infants born between 29 and 43 weeks of gestation. Several networks (visual, auditory, somatosensory, default mode, frontoparietal, and executive control) were examined and appeared to emerge at differing rates, though all showed rapid change during the period from 30 to 40 weeks PMA. Similarly, van den Heuvel et al. [143] imaged 27 PT infants (birth GA 24–29 weeks) at 30 and/or 40 weeks PMA and found evidence of resting state networks by 30 weeks PMA. He and Parikh [144] also observed significant increases in functional connectivity strength in a large set of resting-state networks between 32 and 52 weeks PMA in a sample of 34 VPT infants (birth GA ≤ 32 weeks). Prematurity is associated with widespread alterations in functional connectivity (see [145‐147] for comprehensive review). Findings reveal impaired cerebral lateralization [148, 149], atypical thalamocortical connectivity [150], and reduced connectivity between “rich” nodes; a set of cortical regions which are typically highly connected [151]. These findings are supported by machine learning studies which have used whole-brain functional connectivity to successfully classify infants as either preterm or full-term [152], and even to estimate birth GA [153].

One focus of functional research has been the aforementioned “default mode network,” (DMN) which has been consistently implicated in higher-order social cognitive behaviors such as self-referential processing and self-projection [154]. Fransson et al. [155] collected resting state data around term age from 12 preterm (between 24 and 27 weeks birth GA) and did not find evidence of an infant equivalent of the DMN. It is worth noting, however, that this study did not include a comparison group of FT-born infants. Smyser et al. [156] observed precursors to the DMN in 10 FT but not 53 VPT (birth GA 26–28 weeks) infants, scanned longitudinally from 26 to 40 weeks PMA. A later study [153] included resting state data from 50 PT-born (23–29 weeks birth GA, without moderate-severe brain injury) compared to 50 FT-born infants, all scanned at TEA. Using machine learning algorithms and 214 regions of interest, they were able to distinguish PT and FT-born infants with 84% accuracy. Here, the default mode network contributed to successful categorization, suggesting that preterm birth contributes to altered functional connectivity even in the absence of brain injury. Similar results have been found with PT-born individuals later in life: at 36 months of age, PT-born children show relatively weaker connectivity between resting state networks including the DMN [157]. One recent study also examined functional connectivity in the amygdala of 65 FT (birth GA > 36 weeks) and 57 PT (birth GA < 30 weeks) infants imaged at TEA [158]. Here, functional connectivity patterns in the PT and FT groups were similar to those of older children and adults, though prematurity was associated with reduced magnitude of activation.

Collectively, these imaging findings suggest an array of maturational changes throughout the brain during the second and third trimesters that might be interrupted or altered by preterm birth. These results indicate two potential sources for the unique vulnerability of the PT brain. First, prematurity is associated with an increased risk of specific brain insults such as intraventricular hemorrhage, white matter damage, and hypoxic-ischemic injury. Secondly, and perhaps more critically, the PT brain shows atypicalities in the extensive network connectivity seen by term, and atypical early life experience—even in the form of early exposure to the extra-uterine environment—may result in these connections forming abnormally. Prematurity appears to be associated with altered brain architecture and connectivity in a variety of cortical and subcortical brain regions, including areas and networks traditionally implicated in social-cognitive processing. However, there are few (if any) biomarkers relating specific brain abnormalities to the long-term cognitive, psychiatric, or social outcomes for which preterm infants are at an increased risk.