Development of visual cortex in human neonates is selectively modified by postnatal experience

  1. Mingyang Li
  2. Tingting Liu
  3. Xinyi Xu
  4. Qingqing Wen
  5. Zhiyong Zhao
  6. Xixi Dang
  7. Yi Zhang
  8. Dan Wu  Is a corresponding author
  1. Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, China
  2. Department of Psychology, Zhejiang Sci-Tech University, China
  3. Children's Hospital School of Medicine, Zhejiang University, China
  4. Binjiang Institute of Zhejiang University, China

Abstract

Experience-dependent cortical plasticity is a pivotal process of human brain development and essential for the formation of most cognitive functions. Although studies found that early visual experience could influence the endogenous development of visual cortex in animals, little is known about such impact on human infants. Using the multimodal MRI data from the developing human connectome project, we characterized the early structural and functional maps in the ventral visual cortex and their development during neonatal period. Particularly, we found that postnatal time selectively modulated the cortical thickness in the ventral visual cortex and the functional circuit between bilateral primary visual cortices. But the cortical myelination and functional connections of the high-order visual cortex developed without significant influence of postnatal time in such an early period. The structure–function analysis further revealed that the postnatal time had a direct influence on the development of homotopic connection in area V1, while gestational time had an indirect effect on it through cortical myelination. These findings were further validated in preterm-born infants who had longer postnatal time but shorter gestational time at birth. In short, these data suggested in human newborns that early postnatal time shaped the structural and functional development of the visual cortex in selective and organized patterns.

Editor's evaluation

We believe that this study will make significant contributions to developmental neuroscience and vision science as it is a novel attempt to study the processes that might be innate or genetically wired and those that emerge due to worldly experiences within the sensory systems. The authors suggest that early postnatal experience and time spent inside the womb differentially shape the structural and functional development of the visual cortex. The use of large neonatal dataset from the developmental Human Connectome Project is impressive and strengthens the claims made in the paper.

https://doi.org/10.7554/eLife.78733.sa0

Introduction

A fundamental question in neuroscience is about the role of experience in neurodevelopment (Arcaro and Livingstone, 2021; Barlow, 1975; Holtmaat and Svoboda, 2009; Nithianantharajah and Hannan, 2006). Vision, given its ecological universality and importance across species, has long been taken as a representative modality to investigate such question (Barlow, 1975; Crair et al., 1998; Gödecke and Bonhoeffer, 1996; Li et al., 2008; Li et al., 2006; Roy et al., 2020), and a framework consisting of two distinct phases was proposed to describe the development of visual cortex (Barlow, 1975; Li et al., 2006; White and Fitzpatrick, 2007). This framework includes an early, experience-independent phase in which the basic layout of neural map is established, and a subsequently experience-dependent phase in which the visual experience refines and shapes the initial neural map. Recent studies further revealed the importance of visual experience during the early period of cortical development, suggesting the intricate interaction between early sensory experience and the endogenous mechanisms in neurodevelopment (Li et al., 2008; Li et al., 2006; Roy et al., 2020). However, these studies were carried out on the model animals (e.g., cat and ferret) using electrophysiological techniques. Due to the lack of noninvasive methods to probe the human infant brain, the influence of early postnatal experience on the development of visual cortex in human infants remains unclear.

Neuroimaging techniques, especially magnetic resonance imaging (MRI), provided an ideal tool to noninvasively measure both the brain structure and function, which, however, remains challenging for the infants’ brain due to subject motion, limited resolution, and difficulty in patient recruitment (Cordero-Grande et al., 2018). Owing to the recent technical advances in both images acquisition and processing methods (Hughes et al., 2017), MRI studies of the perinatal brain, for example, the developing human connectome project (dHCP; http://www.developingconnectome.org/) was initiated to investigate the early structural and functional development of human cerebral cortex. Herein we collected a dataset of multimodal MRI data of 407 neonatal subjects from dHCP to address the above question. The different time intervals between gestational age (GA) at birth and the postmenstrual age (PMA) at scan indicated the difference in the postnatal experience across neonates, which reflects the individual variation of the visual experience. Thus, this time window provides us the opportunity to investigate the contributions of early visual experience versus endogenous neurodevelopment on the development of visual cortex in human newborns.

The main purpose of this study is to describe the early structural and functional development of the ventral visual cortex in human newborns and estimate the contribution of postnatal time (PT) to this process. Particularly, the cortical thickness (CT, Lyall et al., 2015; Sowell et al., 2004) and T1w/T2w-based cortical myelination (CM, Glasser and Van Essen, 2011; Soun et al., 2017) data from dHCP was used to measure the development of cortical morphology and microstructure, respectively. The resting-state functional MRI (r-fMRI, Fitzgibbon et al., 2020) data was used to evaluate the development of cortical circuits based on functional connectivity between cortical areas. We focused on the primary visual cortex (V1) and higher-level visual cortex, namely, the ventral occipital temporal cortex (VOTC, Bi et al., 2016; Grill-Spector and Weiner, 2014). The VOTC contains function-specific regions for biologically important categories such as faces (Kanwisher et al., 1998), bodies (Downing et al., 2001), and scenes (Epstein and Kanwisher, 1998), making this region a critical component in the ventral pathway of visual processing (Bi et al., 2016; Grill-Spector and Malach, 2004; Grill-Spector and Weiner, 2014; Kanwisher, 2010), and the impact of early visual experience on those regions is not fully understood.

Previous studies have described the developmental trajectory of CT and CM in human infants and found a generally increasing trend for both two measurements with PMA in whole brain (Bozek et al., 2018; Fenchel et al., 2020). But the contribution of PT to the development was not clarified because PMA reflects both prenatal and postnatal factors. Using r-fMRI, studies found that the primate newborns at a few days of age already had a proto-organization in visual system (Arcaro and Livingstone, 2017) and human infants within a few weeks of age showed a category-specific network (Kamps et al., 2020). However, the small sample size of those studies might undermine reliability of the results given the well-known instability of r-fMRI signals (Poldrack et al., 2017). Moreover, the human infants participating in a previous study (Kamps et al., 2020) were around 1 month old (mean age: 27 days; range from 6 to 57 days), who might already acquire some visual experience, and thus this study could not exclude postnatal visual experience on the innate functional connectivity.

In this study, we first characterized the general development of structural morphology (CT) and microstructure (CM) in the ventral cortex and estimate the contributions of prenatal time and PT on the structural development. Then, we used r-fMRI to characterize the innate organization of the ventral cortex directly after birth and investigated the effect of PT on functional networks of V1 and VOTC areas. Furthermore, we carried out a mediation analysis to investigate the relationship between functional and structural development in area V1. Finally, we evaluated the structural and functional differences of visual cortex between the term- and preterm-born babies as the latter group had shorter GA at birth but longer PT compared to the former group with equivalent PMA.

Results

General development of cortical thickness and cortical myelination in human infants

To show the general developmental trajectories of CT and CM in the visual cortex during the human neonatal period, we included 407 neonates (187 females, PMA = 41.11 ± 1.73 weeks at scan) from a total dataset of 783 subjects after excluding data that did not pass quality control or did not satisfy inclusion criteria (‘Materials and methods’). In total, 355 of them were term-born (PMA = 41.14 ± 1.70, range from 37.43 to 44.71 weeks at scan; Figure 1a). The general trend and spatial variation of visual cortex development were described within an anatomical mask of ventral cortex, which was then segmented into 34 regions of interest (ROIs) per hemisphere according to the HCP-MMP atlas (Glasser et al., 2016), including the early visual cortex (e.g., V1 and V2), higher-level visual cortex (e.g., VOTC), and anterior part of temporal cortex (Figure 1—figure supplement 1, Figure 1—source data 1). In general, the CT and CM of the ventral cortex significantly increased between 37 and 45 weeks of PMA (r = 0.40–0.49, p<10–9; Figure 1b–d). In addition, we found distinct spatial variation along ventral cortex, for example, posterior–anterior and medial–lateral directions (Figure 1—figure supplement 2a and b). Generally, both CT and CM showed higher correlation with PMA in the posterior than anterior region (r = –0.8 and –0.83; p<0.001), and higher correlation in the medial than lateral part within the ventral visual cortex (r = 0.7 and 0.91; p<0.001; Figure 1—figure supplement 2c and d).

Figure 1 with 4 supplements see all
The development of cortical structural properties in human newborns.

(a) The distribution of neonatal postmenstrual age (PMA) at scan in the study population. (b) The averaged cortical thickness (CT) and (c) the averaged cortical myelination (CM) from 38 to 44 weeks of PMA in the ventral cortex. (d) The correlation between CT/CM and PMA in right (peach) and left (blue) ventral cortex (***p<0.001).

Contribution of postnatal time and gestational age on the development of cortical thickness and cortical myelination

Similar developmental trajectories of CT and CM in the ventral cortex across PMA were observed in the above results. It would be interesting to understand the differential influences of GA (the time from fertilization to birth) and PT (the time from birth to MRI scan) on cortical development because they reflected two very factors (e.g., innate growth versus postnatal experience). The GA of our study population was 39.93 ± 1.26 weeks and the PT was 1.21 ± 1.25 weeks, and the correlation between them was not significant (r = –0.08, p>0.1; Figure 2—figure supplement 1). The results showed that the GA was significantly related to the CM (r = 0.53 and 0.51 for right and left hemisphere, respectively; p<10–9) and CT (r = 0.16 and 0.14; p<0.05) averaged in the whole ventral mask. In terms of regional correlation, the CM was significantly correlated with GA in all 68 ROIs (r > 0.28, false discovery rate [FDR]-corrected q < 0.05), while CT showed significant correlation only in 23 ROIs (r > 0.13, FDR-corrected q < 0.05) (Figure 2a and b). The PT was significantly correlated with CT (r = 0.53 and 0.51 for left and right ventral cortex, p<10–9) for left and right ventral cortex, with 61 of the 68 ROIs showing significant positive correlation (all r > 0.11, FDR-corrected q < 0.05). In contrast, the correlation between PT and CM was overall insignificant in the ventral cortex (r = 0.02 and 0.01, p>0.6) with only two ROIs in the anterior temporal areas showing significant effects (Figure 2c and d). We applied a linear mixed-effect model to test whether the CT (or CM) of the whole ventral cortex was differently influenced by the GA versus PT and found that the GA had a significantly stronger effect on the CM than PT (interaction between GA and PT, p<0.05) but no significant difference was found on the development of CT (p>0.6).

Figure 2 with 2 supplements see all
Contribution of postnatal time and gestational age on the development of cortical thickness and cortical myelination.

The correlation maps between cortical thickness (CT)/cortical myelination (CM) and gestational age (GA) (a, b) or postnatal time (PT) (c, d) in ventral cortex. (e–h) Validation analysis using two subsamples: the first subsample included infants who underwent the scans within 3 days after birth to control for the postnatal time (e, f) and the second subsample included infants whose GA ranged from 39 to 40 weeks to control for the prenatal time (g, h). Note: correlation coefficients in the nongray areas were significant after false discovery rate (FDR) correlation (FDR q < 0.05); * p<0.05, **p<0.05, ***p<0.001 by Pearson correlation.

Furthermore, two subsamples of the dataset were selected to control the effect of one age on the other one. The first subsample included infants who underwent the scans within 3 days after birth (n = 173) to control for PT, and the second subsample included infants whose GA was within a short range from 39 to 40 weeks (n = 100) to control for GA. The patterns in the two subsets were similar to the above results (Figure 2e–h). Taken together, those results suggested that the development of CM heavily depends on the endogenous mechanisms and longer prenatal time would be beneficial for the development of myelination. In contrast, the development of CT in neonates was driven by both postnatal and prenatal time.

Next, we looked into the spatial variability within the ventral cortex and focused on two typical visual areas including the V1 and VOTC (15 ROIs) based on HCP-MMP atlas (Figure 3a, Figure 1—figure supplement 1, Figure 1—source data 1). Both regions showed general increase of CT and CM with PMA (r = 0.23–0.61, p<10–6) but were differently influenced by GA and PT. Particularly, the development of CT was significantly modulated by both the PT (r = 0.26–0.52, p<10–6) and GA (r = 0.18–0.29, p<10–3; except for the left VOTC, r = 0.01; Figure 3a–c) for both regions. But the development of CM was significantly influenced by GA (r = 0.43–0.51, p<10–6) but not PT (r = 0.02–0.09, p>0.05; except for the left V1, r = 0.12, p=0.03). Moreover, we applied a linear mixed-effect model to test the developmental difference of the cortical structure between GA versus PT. The results showed that the CT in two regions had nonsignificantly different influences between GA and PT (p>0.3), but CM showed a significantly different impact from the two factors in V1 (p<0.01) and marginal significance in VOTC (p<0.09).

The development of cortical thickness (CT) and cortical myelination (CM) in primary visual cortex (V1) and higher-level visual cortex (VOTC) in human newborns.

(a) Definitions of V1 and VOTC. Correlation between CT (b, c) or CM (d, e) and gestational age (GA) or postnatal time (PT) in V1 and VOTC. *p<0.05, ***p<0.001.

Innate functional organization of ventral cortex in human infants

Beyond structural development of the neonatal cortex, we further asked how the functional connectivity between the different visual subregions changes during early development. To do so, we first estimated the initial functional connectivity pattern in ventral cortex without the influence of postnatal visual experience in a subsample of subjects who underwent the scans within the first day after birth (n = 73). Homotopic correlation was calculated to reflect the functional distinction in neonatal ventral cortex (Arcaro and Livingstone, 2017; Vincent et al., 2007). The homotopic connections in all ROIs of ventral cortex were significant (mean r = 0.13–0.43, t > 12.87, p<10–9; Figure 4a and b) and were significantly higher than adjacent connections (0.29 ± 0.12 vs. 0.19 ± 0.10, Wilcoxon signed-rank test on the Fisher-Z-transformed r-value: z = 16.32, p<10–9) and distal connections (0.04 ± 0.06, z = 16.32, p<10–9; Figure 4c), suggesting that the arealization (homotopic connection) of ventral cortex already existed at birth.

Figure 4 with 4 supplements see all
The innate functional organization of ventral cortex within 1 day after birth and its development in the first month of age in human newborns.

(a) The pairwise correlation matrix describes the functional correlations among 34 regions of interest (ROIs) across hemispheres in ventral cortex. (b) The scatter graph illustrates the correlation coefficients (r) between 34 pairs of bilateral homotopic areas. (c) The box graph depicts the comparison between homotopic, adjacent, and distant connections at birth. (d) Multidimensional scaling and community groups obtained from the pairwise connections between 34 ipsilateral ROIs in right ventral cortex and the corresponding projection on the cortical surface. The color of the nodes in (d) indicates the cluster identity in the community structure analysis, and the color of the lines connecting the nodes indicates the functional correlation (r value) between two nodes. The correlation coefficient maps between the homotopic correlation and postmenstrual age (e), gestational age, (f) or postnatal time (g) in 34 ROIs are mapped onto the cortical surface, and the nongray areas indicated the significant correlations after false discovery rate (FDR) correlation (FDR q < 0.05). ***p<0.001.

In addition, multidimensional scaling (MDS) analysis of the correlations between 34 ROIs within the right ventral cortex revealed the functional relationship among those areas in neonates within 1 day of life (Figure 4d). Using community structure analysis on the network matrix (threshold r = 0.15, p<10–8; Arcaro and Livingstone, 2017), we further partitioned these areas into three groups, including a lateral cluster, a medial cluster, and an anterior–lateral temporal cluster (Figure 4d). Similar results were found with different threshold (e.g., r = 0.10, 0.15, and 0.20) and in the left hemisphere (Figure 4—figure supplement 1). Taken together, those results suggested that the proto-organization of ventral visual cortex was formed before acquiring any higher-level visual experience.

Development of functional connections in the ventral visual cortex

Based on the baseline functional connections in term-born neonates, we were intrigued to know how those functional properties of the ventral visual cortex develop during neonatal stage of human infants? If so, whether PT and GA had different influences on this process? For the arealization in the 34 ROIs, V1 and lateral–anterior regions showed a significant increase with PMA (16/34 regions, r = 0.12–0.27; FDR q < 0.05; Figure 4e), suggesting a general development of homotopic connection in the ventral cortex. However, the GA showed significant correlations with the homotopic connections in the anterior temporal cortex (14/34 patches, r = 0.12–0.28; FDR q < 0.05; Figure 4f) but PT was significantly related to the connections in V1, fusiform, and anterior temporal areas (6/34 patches, r = 0.14–0.22; FDR q < 0.05; Figure 4g). For the entire ventral cortex, we applied MDS and community structure analysis in each PMA week from 38 to 44 weeks across all term-born infants (n = 355) and found similar three-cluster network structures in the infants with different PMA (except for 42 weeks; Figure 4—figure supplement 2). Then we used two typical network measurements, including global efficiency and mean cluster coefficient, to quantify the development of the functional network in ventral cortex. Both two measurements showed significant increase with PMA (r = 0.25 and 0.27, p<0.001) and significant correlations with both GA (r = 0.11 and 0.13, p<0.05) and PT (r = 0.22–0.25, p<0.001; Figure 4—figure supplement 3).

The correlations between ipsilateral connections and PMA showed both positive and negative results in different regions of the ventral cortex. Particularly, the early visual cortex (e.g., V1 and V2) showed decreasing connection to other visual areas across PMA while some areas in the VOTC (e.g., fusiform and parahippocampus) showed increasing tendency (Figure 4—figure supplement 4). Here, we focused on the early connections within the V1 and VOTC (Figure 5a) and found the homotopic correlation between contralateral V1’s (r = 0.43 ± 0.25) was significantly higher than the correlation between VOTCs (r = 0.38 ± 0.22; t = 4.29, p<10–4), which were both higher than the connections between V1-VOTC within the same hemisphere (r = 0.16 ± 0.19; t = 15.74 and 13.49, p<10–9; Figure 5b) in the term-born infants. Furthermore, the homotopic correlations of both V1 and VOTC showed significant increase with PMA (r = 0.16 and 0.18, p<0.01) but the connection between V1 and VOTC showed significant decrease with PMA (r = –0.23, p<10–4). Moreover, the homotopic correlations of V1 and VOTC were significantly modulated by the PT (r = –0.20 and –0.11, p<0.05) but not the GA (r = –0.01 and 0.09, p>0.05; Figure 5c and d). The ipsilateral connection between V1 and VOTC was significantly correlated with both PT and GA (r = –0.19 and –0.13, p<0.05; Figure 5e).

The development of the functional connections between bilateral primary visual cortex (V1) and higher-level visual cortex (VOTC) and their cross-correlations.

(a) The connections of interest include the homotopic connection between bilateral V1 (purple), bilateral VOTC (peach), and the averaged ipsilateral connections between V1 and VOTC in each of the hemispheres (gray). (b) Comparison of the three types of connections in the infants within 1 day after birth (no postnatal experience). The correlation between gestational age (GA) or postnatal time (PT) and bilateral V1 connection (c), bilateral VOTC connection (d), and ipsilateral connection between V1 and VOTC (e). The values in the bracket indicate the partial correlation coefficient controlling for postmenstrual age (PMA) of infants. Contra, contralateral; Ipsi, ipsilateral; *p<0.05; **p<0.01, ***p<0.001.

Relationship between structural and functional properties in the area V1

The above results revealed that structural and functional properties of the ventral visual cortex both developed with PMA, but were differently influenced by the in utero and external environment (Table 1). We further investigated the relationship between structural and functional development in area V1, which showed a strong developmental effect in both structural and functional analyses. Mediation analysis was employed to see whether the development (GA or PT) of the homotopic connection between bilateral V1 was mediated by the structural properties (CT or CM). We found that the PT had a significant direct effect on the homotopic function that was not mediated by CT or CM (Figure 6a and b). In contrast, the direct effect of GA on the homotopic connection was not significant but the indirect effect of GA through CM on the connection was significant (Figure 6c and d).

Table 1
Summary of the effects of gestational age (GA) and postnatal time (PT) on the structural and functional properties of the ventral visual cortex.

The numbers indicate the correlation coefficients of the corresponding measurement and the GA or PT.

PropertiesGAPT
 CT Whole ventral cortex<0.2>0.4
 V10.2–0.4>0.4
 VOTC<0.20.2–0.4
 CM Whole ventral cortex>0.4Nonsignificant
 V1>0.4<0.2
 VOTC>0.4Nonsignificant
 FC network properties in the whole ventral cortex0.2–0.4<0.2
 Homotopic connection V1Nonsignificant<0.2
 VOTCNonsignificant<0.2
 Ipsilateral connection between V1 and VOTC<0.2<0.2
  1. CT, cortical thickness; CM, cortical myelination; VOTC, ventral occipital temporal cortex.

Mediation analysis between the developmental factors (gestational age [GA] or postnatal time [PT]), homotopic functional connection between bilateral V1, and structural features (cortical thickness [CT] or cortical myelination [CM]).

Homotopic connection between bilateral V1 was set as an independent variate, while the developmental factor was a dependent variate, and the structural features (CT or CM) was the mediated variate. **p<0.01, ***p<0.001.

Comparison of structural and functional features between term- and preterm-born infants

Compared the term-born infants, preterm-born infants have longer PT with relatively shorter GA at birth, and thus the comparison between them may help us to understand the influences of early experience and innate growth on cortical development. Considering the unbalanced sample size between the two groups (n = 355 vs. 52), we selected a subgroup in the term-born infants with equal sample size (n = 52) and similar PMA to the preterm-born neonates (40.92 ± 1.87 vs. 40.90 ± 1.91 weeks; t = 0.06, p>0.95) (Figure 7a). The term-born infants had significantly higher GA (39.96 ± 1.40 vs. 31.98 ± 3.32 weeks; t = 14.04, p<10–9) but lower PT (0.96 ± 1.13 vs. 8.91 ± 4.42 weeks; t = –13.86, p<10–9) compared to preterm-born infants. For the structural measurements, the mean CT across all ROIs was significantly lower in the term-born (1.39 ± 0.05) than preterm-born infants (1.35 ± 0.05; t = –3.16, p<0.01) and 10 of 68 ROIs showed significantly lower CT in the term-born than preterm-born infants (Figure 7—figure supplement 1). On the contrary, the mean CM across all ROIs was higher in the term-born (1.22 ± 0.10) than preterm-born (1.09 ± 0.08; t = 7.11, p<10–9) infants and all ROIs were significantly higher in the term-born than preterm-born groups. Particularly, the term-born infants showed significantly lower CT in the area V1 (t = –2.55, p<0.02) but not VOTC (t = –1.24, p>0.2; Figure 7f) and higher CM in both area V1 and VOTC (t = 6.65 and 7.54, p<10–8; Figure 7g).

Figure 7 with 2 supplements see all
Comparison of structural and functional features between term- and preterm-born infants.

(a) The comparison of gestational age (GA), postmenstrual age (PMA), and postnatal time (PT) between term-born and preterm-born infants. (b–d) The t maps of cortical thickness (CT), cortical myelination (CM), and arealization between two groups (term – preterm) in ventral visual cortex. Only the significantly different regions (false discovery rate [FDR] q < 0.05) are shown. (e–h) The box plots of structural and functional measurements between two groups, including the global coefficient and mean cluster coefficient of the ventral network (e), CT in V1 and VOTC (f), CM in V1 and VOTC (g), and functional connectivity between contralateral V1 and VOTC (h). *p<0.05, **p<0.01, ***p<0.001 by two-sample t-test.

For the functional organization of ventral cortex, two groups showed similar three-community structure except that some areas (e.g., V1 and V2; Figure 7—figure supplement 2) were included in adjacent community in the preterm-born group. Significantly different network measures were found between two groups, for example, the term-born infants had higher global efficiency and mean cluster coefficient than preterm-born infants (t = 2.22 and 2.43, p<0.05; Figure 6e), suggesting higher efficiency of information transmission and higher differentiation of cortical function in the ventral cortex of term-born infants. Compared to preterm-born infants, the arealization in the term-born infants was higher in posterior and middle parts of ventral cortex but lower in the anterior temporal areas (Figure 6d). Particularly, the preterm-born infants showed lower arealization in the VOTC (t = 3.75, p<0.001) but not V1 (t = 1.60, p>0.1) than term-born neonates (Figure 6h), suggesting that the connections between bilateral visual cortex were also influenced by the GA.

Discussion

Experience-dependent plasticity is one of the most striking features of human brain and also the foundation of acquired cognitive ability. Utilizing the large datasets of neonatal multimodal MRI images from dHCP, we reported the early structural and functional maps in the ventral visual cortex and their development during the first month of age, focusing on the contributions of prenatal time and PT to the cortical development in this early period. Particularly, we found that PT was significantly related to the cortical thickening and specific functional circuits (e.g., homotopic functional connections between the bilateral visual cortex), while GA was significantly correlated with the CT, CM, and ipsilateral connections in the visual cortex. Such separation reflected the influence of the developmental environment (in utero and outside world) on the cortical development in the infants. Furthermore, we found that the preterm-born infants had higher CT (e.g., V1), lower CM, and lower homotopic connections (e.g., VOTC) compared to term-born infants at equivalent PMA. Taken together, those results suggested that the development of human visual cortex does not strictly follow the two-stage hypothesis (Barlow, 1975; Li et al., 2006; White and Fitzpatrick, 2007) as the cortical properties are modulated by postnatal environment even within the first month of life.

Similar to the previous findings in human infants (Bozek et al., 2018; Fenchel et al., 2020), both the CT and CM in ventral cortex increased between 37 and 44 weeks of PMA in our study. We went one step further to investigate the different mechanisms underlying these trends and found the development of CT was considerably modulated by both postnatal and prenatal times while the CM was heavily influenced by prenatal duration. CT is related to the synaptogenesis and synaptic pruning processes that are shaped by both innate and postnatal factors. It was reported that human visual cortex thickens to a maximum during the first 2 years and gradually thins thereafter (Lyall et al., 2015), mirroring the trajectory of synapses in human striate cortex (Huttenlocher and de Courten, 1987). The previous evidence that both congenitally blind and sighted subjects showed uptrend of CT in the early stage, leading to the conclusion that the visual experience might not be involved in this process (Bourgeois and Rakic, 1996; Jiang et al., 2009). Our results in the human neonates suggested the postnatal environment could accelerate the thickening of neonatal visual cortex, even if it was not necessary.

Is it the visual experience out of various postnatal stimuli that modulated the development of CT in ventral visual cortex? Although we could not provide a direct evidence from this study, there were indications that the CT in this study might reflect an experience-dependent synaptogenesis induced by early visual experience (Holtmaat and Svoboda, 2009). On the one hand, we found that across whole brain, PT-dependent increase of CT was most prominent in the V1, primary auditory area, and central sulcus (Figure 2—figure supplement 2), which directly receive visual, auditory, sensorimotor stimuli; while the increases of CM were mainly distributed in the frontal areas (Figure 2—figure supplement 2). This whole-brain pattern suggested that CT was shaped by postnatal environment universally, and it is reasonable to expect that ventral visual cortex is closely related to visual experience. On the other hand, we found a well-organized spatial pattern of cortical thickening in ventral cortex with the posterior area showing faster development than anterior regions. This might reflect the influence of visual flow along the ventral cortex, for example, the primary visual area in the posterior area processes all kinds of visual information while higher-level visual cortex responds less. Future study could combine the behavioral measurements to test the relation between these two factors or compare the sighted and congenitally blind infants to further illustrate such question.

T1w/T2w-based myelination measurement, which takes advantage of the covariation between cortical myelin content and T1w (positive covariation), as well as T2w (negative covariation) intensities, could be used as a surrogate maker of myelination (Glasser and Van Essen, 2011; Soun et al., 2017). CM is related to various cognitive performance (Glasser et al., 2014), which might reflect the increase of myelinated fibers in the cortex. Previous studies reported the coupling between CM and face processing, suggesting an experience-dependent development of visual cortical microstructure in later life (e.g., 5–12 years old; Nordt et al., 2021). However, we found a significant correlation between CM in the ventral cortex and GA but not PT in the neonatal period, suggesting that early myelination was primarily determined by the endogenic growth in the uterine environment. The posterior-to-anterior spatial pattern of CM in the ventral visual cortex agreed with a previous study on infants at around 6 months of age (Natu et al., 2021). Our results further suggested that this hierarchical pattern occurs as early as the neonatal period.

Using r-fMRI data, previous studies evaluated the functional organization of the visual system in neonatal macaques (Arcaro and Livingstone, 2017) and category-selective network in human infants (6–57 days) (Kamps et al., 2020). Both studies revealed the proto-organization prior to the emergence of functional domain such as face and place areas. Herein we validated those findings in a large sample with infants as early as 1 day of postnatal age, fully controlling the influence of postnatal visual experience. Particularly, we found two distinct visual clusters along the medial–lateral axis, which captured the central–peripheral organization of ventral visual system (Grill-Spector and Weiner, 2014; Hasson et al., 2002; Levy et al., 2001; Wiesel, 1982). Apart from the border of two medial–lateral clusters, the lateral area becomes more face-selective while the medial area is specialized in processing scene and buildings later in life (Nordt et al., 2021), and thereby our observation suggests the innate scaffold is already established at birth for subsequent experience-dependent modification in ventral visual cortex (Arcaro et al., 2017; Arcaro and Livingstone, 2021; Arcaro and Livingstone, 2017). However, the present homotopic connections in the human neonates were lower than those in neonate macaca mulattas (Arcaro and Livingstone, 2017). This difference might relate to the higher motion in human infants, less r-fMRI data in this study, coarser parcellation in the visual cortex used in this work, and the developmental difference between primates and humans in the neonatal period. Furthermore, the connections among the ventral visual cortex have developed during this early stage. Specifically, the homotopic connections between bilateral V1 and between bilateral VOTC both increased with PMA, indicating increased degree of functional distinction (Arcaro and Livingstone, 2017; Vincent et al., 2007). In contrast, the connection between V1 and VOTC decreased between 37 and 44 weeks of PMA, which might also reflect the development of functional differentiation between adjacent regions in the same hemisphere. More importantly, the connection between bilateral V1 and VOTC was significantly modified by PT, supporting that the early postnatal experience not only influences the local structural features of cortical cortex but also the functional circuits. It is worth noting that the increased homotopic connection can be direct or indirect, for example, the effect might be driven by external regions with enhanced connection to both of the areas (e.g., thalamus).

The preterm-born babies have longer PT but shorter GA compared to full-term infants at the same PMA. The findings that CT in the ventral cortex was generally lower in the term-born than preterm-born infants and that CM showed an opposite trend between the two groups supported the above analysis that CT was preferably influenced by PT while CM was largely dependent on GA during the neonatal period. In terms of functional development, we found the homotopic connections were lower in preterm than in term-born infants, in contrast to the above finding that the homotopic connection increased with PT in both V1 and VOTC. These results might be due to the mediation effect of CM on the connection.

In brief, our results suggested that early cortical development is a mixed outcome of endogenous and experience-dependent development in both cortical structure and function, and postnatal experience could selectively modify the endogenous development of visual cortex during early infancy.

Limitation

One limitation of this study is the comparison between preterm- and term-born infants did not consider the different visual experience in these infants. The preterm-born neonates may experience very different environment than those of the term-born, for example, the preterm environment can be heavily regulated if they were in a NICU, but we did not have detailed information about the postnatal environment to control for it. Meantime, both GA and PT were different between preterm- and term-born neonates. Then any of the differences between the two groups might have come from the combined effects of GA and PT, and unfortunately, we were not able to separate them in this study. Another concern was the partial-volume effect on the cortical measurements. The changing thickness of cortical ribbon during development may change the degree of partial-volume effect, and thus may affect the CM measurement and may contribute to the myelination difference observed between preterm- and term-born groups.

Materials and methods

Participants

A total of 887 datasets (783 neonatal subjects) from dHCP (http://www.developingconnectome.org/) were collected. The dHCP study was approved by the UK Health Research Authority (14/LO/1169), and written consent was obtained from the parents or legal guardians of the study subjects. We excluded the subjects who (1) had high radiology score (above two points) reviewed by specialist perinatal neuroradiologist, which indicated possibly clinical or analytical significance (e.g., punctate lesions or other focal white matter or cortical lesions but not considered to be of clinical significance; https://biomedia.github.io/dHCP-release-notes/structure.html) (n = 218, including two datasets without scores); (2) were sedated during scan (n = 5); (3) were scanned early than 37 weeks (n = 91) of PMA; (4) missing myelin maps or r-fMRI data (n = 152); (5) could not pass the quality control of dHCP preprocessing pipelines for the structural (n = 1) or r-fMRI data (n = 9); and (6) failed the cortical registration pipeline (n = 4). Finally, 407 datasets (407 subjects) were selected in this study, in which 355 were term-born (GA: 39.93 ± 1.26 [37–42.29] weeks; PMA at scan: 41.14 ± 1.7 [37.43–44.71] weeks) and 52 were preterm-born (GA: 31.98 ± 3.35 [23.71–36.86] weeks; PMA at scan: 40.9 ± 1.91 [37–44.29] weeks).

Data acquisition

Request a detailed protocol

All neuroimaging data were acquired at the Evelina Newborn Imaging Centre, Evelina London Children’s Hospital, using a 3 T Philips scanner with a newly designed neonatal imaging system including a customized 32-channel phased array head coil, an elaborated positioning device, and a custom-made acoustic hood for infant (Hughes et al., 2017). Infants were imaged during natural sleep without sedation. Structural (T1 and T2 weighted image), r-fMRI and diffusion MRI (dMRI) were collected within a single-scan session for each neonate over 63 min. T2-weighted (T2w) and inversion recovery T1-weighted (T1w) multi-slice fast spin-echo images were acquired in sagittal and axial stacks with in-plane resolution of 0.8 × 0.8 mm and slice thickness of 1.6 mm (0.8 mm overlap, except for sagittal T1w that used 0.74 mm). Other parameters were as follows: (1) T1w images were acquired with repetition time (TR) = 4795 ms, echo time (TE) = 8.7 ms, inversion time (TI) = 1740 ms, SENSE factor of 2.27 (axial) and 2.66 (sagittal), and filed of view (FOV) = 145 × 122 × 100 mm; and (2) T2w images were acquired with TR = 12,000 ms, TE = 156 ms, SENSE factor of 2.11 (axial) and 2.58 (sagittal), and FOV = 145 × 145 × 108 mm. The axial and sagittal stacks were integrated using a super-resolution method (Kuklisova-Murgasova et al., 2012). High temporal resolution r-fMRI optimized for neonates were collected using echo-planar imaging with a multiband factor of 9 in 15.05 min: TR = 392 ms, TE = 38 ms, 2300 volumes, flip angle = 34°, and spatial resolution = 2.15 mm isotropic (https://biomedia.github.io/dHCP-release-notes/acquire.html; Bozek et al., 2018; Fitzgibbon et al., 2020).

Data preprocessing

Request a detailed protocol

We collected the preprocessed anatomical and r-fMRI data from dHCP database (Makropoulos et al., 2018). For the anatomical data, briefly, preprocessing of dHCP pipeline included super-resolution reconstruction to obtain the 3D T1w/T2w volumes (Kuklisova-Murgasova et al., 2012), registration (from T1w to T2w), bias correction, brain extraction, segmentation (on T2w volume using DRAW-EM method, Makropoulos et al., 2014; see Figure 1—figure supplement 3), surface extraction (Schuh et al., 2017), and surface registration (Robinson et al., 2018). The initial structural data from dHCP in this study included the individual brain surfaces, CT (corrected version for uneven vertex sampling of gyri relative to sulci), and T1w-/T2w-based myelination maps. Surfaces and cortical metrics of each neonate were nonlinearly aligned to the dHCP symmetric template (40 weeks of PMA; Bozek et al., 2018) using multimodal surface matching (MSM, Robinson et al., 2018).

The collected functional data from dHCP included the r-fMRI data in individual space and the motion parameters (Fitzgibbon et al., 2020). These images were further preprocessed with the following steps using custom codes and DEPABI toolbox (Yan et al., 2016) in MATLAB (v2018a). (1) A conservative approach was adopted to address the severe head motion in infants. Similar to the previous study (Eyre et al., 2021), we selected a continuous subset (1600 volumes, around 70%) of the data with lowest head motion for each of the infants. Specifically, we first estimated the variation of head motion for each time point within a 50 volume window by calculating the standard deviation of the head motion within the time window, then we calculated the sum of the variations in all possible 1600 continuous volumes (e.g., 1–1600, 2–1601, …, 701–2300), and finally chose the subset with lowest sum of the head motion variation. (2) Registration of the selected r-fMRI subset to individual T2w space with FLIRT (Jenkinson et al., 2002). (3) Linear detrending. (4) Regression of nuisance covariates, including 24 head motion parameters and the signals of white matter, cerebrospinal fluid, and global brain. These tissue masks were extracted from the T2w-based segmentation of the corresponding subject. (5) Temporal bandpass filtering with a pass band of 0.01–0.08 Hz. (6) The filtered r-fMRI volume was further projected into the individual cortical surface and then registered to the dHCP symmetric template using MSM method. The resulting surface data was slightly smoothed with 2 mm full-width-half-maximum (FWHM) using workbench (Glasser et al., 2013).

Anatomical ROIs

Request a detailed protocol

All ROIs were defined in adult space based on the HCP-MMP atlas (Glasser et al., 2016) and registered into the dHCP neonatal template (40 weeks) using MSM method with cortical sulcus as the features to drive the alignment (Robinson et al., 2018). We then registered all neonatal surfaces onto the dHCP template using the transformation matrices provided by the dHCP pipeline. We visually checked the registration results for some of the typical sulci and gyri in Figure 1—figure supplement 1. Although the best way is to use a neonate-specific parcellation for the present analysis, there is no such fine cortical parcellation of human neonates yet. In addition, we used a definable area – the V1 area – as an example to quantitatively evaluate the registration quality. Specifically, we manually delineated the area V1 on the gradient map (Figure 1—figure supplement 4a) of the averaged CM map and calculated the dice coefficient between the transformed V1 from Glasser atlas and the manually delineated V1. The result showed a dice score of 0.83 and 0.80 for right and left hemispheres, respectively. As shown in Figure 1—figure supplement 4b, the location of the registered V1 and manually delineated V1 area was consistent, although the latter was slightly bigger. We compared the parcellations on the neonatal atlas to those on an adult the S1200 group-average data collected from the human connectome project (HCP; https://www.humanconnectome.org/) and found the parcellations were visually acceptable at large scale (Figure 1—figure supplement 1).

The ventral cortex was parcellated into 34 ROIs per hemisphere in the HCP-MMP atlas, which contain basic visual cortex (e.g., V1 and V2), higher-level visual cortex (e.g., VOTC), and anterior part of ventral temporal cortex (Figure 1—figure supplement 1, Figure 1—source data 1). The VOTC includes 15 ROIs per hemisphere, including common category-selective regions such as parahippocampus, middle fusiform, and lateral occipital cortex (Bi et al., 2016; Grill-Spector and Weiner, 2014). The primary visual area was defined as the V1 ROI in the HCP-MMP atlas (Figure 1—figure supplement 1).

Data analysis

Experimental design

Request a detailed protocol

The purpose of this study is to describe the structural and functional development of ventral visual cortex in human newborns and evaluate the influence of postnatal experience in this process. To depict a comprehensive picture of the spatiotemporal development, for each structural or functional measurement, we first presented the general developmental trend of whole ventral cortex and also the spatial variations. Then we focused on the V1 and VOTC and characterized the detailed developmental patterns of these two areas.

To evaluate the contributions of GA and PT on the development, we first used Pearson correlation between the developmental factors and specific structural properties. In addition, we designed two subsamples to validate the independent influence of GA (or PT). The first subsample included infants who underwent the scans within 3 days after birth and thus had limited postnatal experience, and the second subsample included infants within a limited GA range from 39 to 40 weeks to estimate the effect of prenatal time.

Lastly, we compared the term-born infants with preterm-born peers who had longer postnatal experience at equivalent PMA, which may help to explain the influence of early experience versus endogenous coding on cortical development. Therefore, for the structural and functional measurements that were significantly influenced by PT, we will further compare them between term and preterm-born infants.

Statistical analysis

Request a detailed protocol

For all of the correlation analysis, we removed the data beyond 3 SDs from the mean value across all infants. The correlation coefficients were Fisher-Z-transformed before any statistical test, and the plots (e.g., box plots and scatter plots) used the original r value to give an intuitive picture. Extraction and calculation of cortical measurement (e.g., CT, CM, and r-fMRI) were performed using gifti toolbox (https://www.nitrc.org/projects/gifti/) in MATLAB (v2018a) and commands in workbench (v1.5.0, https://github.com/Washington-University/workbench; Harwell, 2022). Statistical analysis, including Pearson correlation, t-test, and linear mixed-effect model, was performed in MATLAB. Mediation analysis was conducted using the PROCESS toolbox (v4.1) in SPSS (v21). FDR correction was used for the multiple-comparison correction if not specifically mentioned. p<0.05 or corrected q < 0.05 were considered significant for all tests.

Structural data

Request a detailed protocol
Cortical thickness and myelination maps of ventral cortex
Request a detailed protocol

To generate the average cortical measurements (CT and CM) at the group level, we first registered the metrics of every subject onto a common 40-week template surface using the registration files provided by dHCP pipeline, and then averaged them in a vertex-wise manner across subjects within a specific week (e.g., 38-week map included neonates whose PMA ranged from 37.5 to 38.5 weeks). Because there were only two infants in the 37-week PMA group (37–37.5 weeks) and one infant in the 45-week PMA group (44.5–45.5), we did not present the averaged maps of those weeks.

Spatial differences between developmental trajectories
Request a detailed protocol

To quantify the spatial difference in the developmental change of these cortical measurements, we extracted the default coordinate system of the ventral cortex in the atlas, in which the y-axis was along the anterior–posterior direction and the x-axis was along the medial–lateral direction (Figure 1—figure supplement 2b). For each axis, we divided the ventral cortex into 30 segments with equal length along the axis. We then averaged the correlation coefficients across all vertexes within a segment to obtain the spatial pattern of cortical development along the two axes (Figure 1—figure supplement 2b–d).

Comparison of the developmental pattern of cortical structural properties with respect to GA versus PT
Request a detailed protocol

We applied a linear mixed-effect model to test whether structural measurements (CT or CM) were differently influenced by GA versus PT, in which CT or CM was independent variate, GA, PT, and their interaction were fixed effect, and the intercept was the random effect.

R-fMRI data

Temporal correlation analysis

Request a detailed protocol

For each subject, Pearson correlations were carried out on the ROI-averaged time series within and across the left and right ventral cortex. The resulting connections were divided into three groups, namely, the homotopic connection (the connection between two paired areas in two hemispheres. for example, right and left V1), adjacent connection (e.g., right V1 and left V2 since V1 and V2 are adjacent), and distant connections (two areas that were not the paired or adjacent). Independent-samples t-test was used to test whether the homotopic correlation was significantly greater than zero across subjects. To compare the correlation among the three types of connections, we applied a nonparameter statistical analysis (Wilcoxon signed-rank) across subjects.

Multidimensional scaling

Request a detailed protocol

Similar to the method used in the previous studies (Arcaro and Livingstone, 2017), for each hemisphere, we obtained the ipsilateral correlation matrix between the 34 patches in ventral cortex. Then the correlation matrix was transformed into a distance matrix using 1 – r for each entry of the matrix. Nonclassical MDS was carried out on the distance matrix with Kruskal’s normalized criterion in MATLAB. The 34 ROIs were plotted in a two-dimensional coordinate according to the first two principal values in the MDS analysis.

Clustering analysis

Request a detailed protocol

For the correlation matrix between the 34 ROIs within ipsilateral ventral cortex, a threshold of r = 0.15 was used to remove weak and negative connections. Regional clusters in the ventral cortical network were obtained using a spectral algorithm (Newman, 2006) in the Brain Connectivity Toolbox (BCT; Rubinov and Sporns, 2010). Different threshold (e.g., 0.1 and 0.2) was also used to validate the results (Figure 4—figure supplement 1).

Developmental changes of ventral network

Request a detailed protocol

Two typical network measurements, including global efficiency and mean cluster coefficient (calculated in BCT), were used to quantify the integration and segregation of the ventral network. Pearson correlation between two measurements and PMA (or PT) was used to describe the development of the ventral network in human newborns.

Developmental change of the connections in V1 and VOTC

Request a detailed protocol

Four ROIs, including V1 and VOTC in two hemispheres, were included in this analysis. Pearson correlation was performed between the averaged time series in each four ROIs, resulting in two contralateral connections between V1 and VOTC, and two ipsilateral connections between V1 and VOTC in each hemisphere. We then calculated the Pearson correlation coefficient between these connections (Fisher-Z-transformed) and GA (or PT) to measure the developmental changes.

Comparison between term-born and preterm infants

Request a detailed protocol

Due to the unbalanced sample size between preterm-born (n = 52) and term-born infants (n = 355), we selected a subgroup in the term-born infants with equal sample size and similar PMA (40.92 ± 1.87 weeks) to the preterm-born neonates (40.90 ± 1.91 weeks; t = 0.06, p>0.95). Specifically, for each infant in the preterm-born group, we chose a matched term-born subject with the closest PMA. Structural and functional measurements were compared between two groups using two-sample t-test.

Data and code availability

Request a detailed protocol

The data used in this work are available online on the dHCP page (http://www.developingconnectome.org/data-release/third-data-release/).

Main MATLAB codes used for data preprocessing and analysis are available at https://github.com/MingyangLeee/Neonatal-Visual-Cortex. (Li, 2022; copy archived at swh:1:rev:d9e50dc27b6ae577e644b6da4abc2ec6d68b7900).

Appendix 1

Appendix 1—key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Software, algorithmMATLABMathWorksRRID:SCR_001622
Software, algorithmSPMFILRRID:SCR_007037
Software, algorithmDPABIdoi:10.1007/s12021-016-9299-4RRID:SCR_010501
Software, algorithmFSLFMRIBRRID:SCR_002823
Software, algorithmMSMdoi:10.1016/j.neuroimage.2017.10.037
Software, algorithmBCTdoi:10.1016/j.neuroimage.2009.10.003RRID:SCR_004841
Software, algorithmConnectome WorkbenchConnectome WorkbenchRRID:SCR_008750

Data availability

The data used in this work are available online on the dHCP page (http://www.developingconnectome.org/data-release/third-data-release/).

References

    1. Huttenlocher PR
    2. de Courten C
    (1987)
    The development of synapses in striate cortex of man
    Human Neurobiology 6:1–9.

Decision letter

  1. Vaidehi Natu
    Reviewing Editor; Stanford University, United States
  2. Joshua I Gold
    Senior Editor; University of Pennsylvania, United States
  3. Cameron Ellis
    Reviewer; Haskins Laboratories, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Development of visual cortex in human neonates are selectively modified by postnatal experience" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Joshua Gold as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Cameron Ellis (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Overall, the assessment of the 3 reviewers is highly consistent. They believe that this study will make significant contributions to developmental neuroscience and be of broad interest to the developmental neuroscience community. That being said, the reviewers brought up a number of questions and concerns that they would like for you to address before the manuscript can be further evaluated. The reviewers all provided detailed recommendations that you should consider for your revision (refer to "Recommendations for the authors" for individual reviewer clarifications), but here are the main points that we invite you to address in the revision. Naturally, at this point, there is no guarantee that the revision will be accepted for publication. This depends on an evaluation of your responses by the reviewers, as well as any new issues that arise during the process.

Major Concerns:

1. One important concern brought up by all three reviewers is the confusion with the terminologies GA, PMA, and post-natal time (PT), for preemies/full-term infants. Please define and simplify terms as suggested by the reviewers to ameliorate the readability and justify why the authors chose to use each of the terms at different times in the analyses. Overall, the reviewers suggested using GA and PT and replacing the analyses with GA to simplify the findings. Please see individual reviewers' comments to simplify the language and avoid confusion across the manuscript.

2. To make the results and the flow of the paper more cohesive, we recommend the authors frame and justify each analysis with a clear question that the analysis will answer or the hypothesis that the authors posit as it is often difficult to understand how an analysis being described would contribute to the overall claims of the paper.

3. Relatedly, the reviewers also brought up important points in several places whereby the claims made in the paper are not supported by the analyses conducted under this claim (e.g.., in cortical myelination and overall functional connectivity analyses in Figure 2,4, and 5 as well as the final results). The statistical tests and analysis conducted need to support the conclusions and claims. We recommend that the authors rephrase or tone down the conclusions made in this analysis or reassess their analyses to support the claims. A similar concern is raised by reviewers 1 and 3 in the final section "Comparison between structural and functional properties…", as the reader is left expecting a formal comparison between structural and functional development. Please consider an analysis as suggested by reviewer 3 for e.g., a regression with functional connectivity and structural metrics, to ask for example if a region's homotopic connectivity is correlated with CT or CM and which aspect of time (gestation or postnatal) is most important.

4. Given the authors used an adult atlas to draw regions on the infants' brains, a sanity check of how well the Glasser atlas regions align on a neonatal brain at e.g., 37 weeks and compare that to 42 weeks old and adult brain and running some statistics on how these cortical-based alignments from the adult brain to the neonatal brain align, for a few sample subjects is recommended.

5. A similar sanity check is also essential to check the white-gray matter boundary segmentation in some sample infant data as the authors use cortical thickness as one of their measures and CT is measured as the distance between the pial surface and gray-white matter boundary and the cortical ribbon of infants is thin at birth, there may be a possibility that partial-volume effects could be more prevalent in less-developed infants and impact myelin metrics.

6. The reviewers also asked if the authors used Fisher Z to transform their correlations? This is important to clarify as a significant portion of the results in the paper are correlations and it is unclear from the methods whether the authors are transforming their correlation values to make that use appropriate.

7. Since one of the big conclusions of this paper is that not all structural and functional aspects are affected by gestation or postnatal time the same way, having a quick summary of those findings would be helpful. The reviewers suggest a summary table showing the morphological and myelin measures against gestational time and prenatal time, whether the gestational or prenatal time was a significant modulator of that measure.

8. Reviewer 2 brings up an important point about the homotopy analysis that needs clarification and perhaps rethinking of the statistical tests (e.g., comparing 34 homotopic pairs with the hundreds of non-homotopic pairs) that must be addressed in the revision. Please also consider the recommended alternative test by reviewer 2 that would compensate for some of the constraints to measure the homotopic correlation for a region.

Reviewer #1 (Recommendations for the authors):

While I find this work theoretically well-motivated and the use of the large dHCP dataset very exciting, there are some major concerns that must be addressed before this study is ready for publication:

1. There was a lot of confusion for me about the terminology between GA, PMA, and post-natal time, for preemies/full-term infants. To simplify the matter, I would first define these ages for the reader and also choose to use a single term, preferably GA as postmenstrual age would be GA plus the chronological age. This would make it easy to follow the flow of the analysis, especially in the section: "Contribution of postnatal time on the development of cortical thickness and cortical myelination". I also feel the same about the analysis in figure 3, it would make it clearer to use GA and postnatal time, as PMA already included both and it's difficult to see the relationship of the GA on CT and CM separately.

2. Relatedly, I really like Supplementary Figure 4c and 4f, and they may be your point on the postnatal age and GA influence on CT and CM really clear and they can be moved as one of the main figures in the paper.

3. It would be good to sanity check how the adult Glasser maps align on a neonatal brain at e.g., 37 weeks and compare that to 42 weeks old and adult and run some statistics on how these cortical-based alignments from the adult brain to the neonatal brain align, for a few sample subjects. Another sanity check that is necessary especially with the cortical thickness measure is to check the segmentations in the data. It would be useful to provide a montage of a few sample subjects showing the segmentation of the gray-white boundary as the contrast properties in infants is hard to decipher.

4. In the first analysis are there any hemispheric differences in the relationship between CT and CM and age?

5. Can the authors describe how the average cortical thickness and myelination maps in Figures1 b, c are generated? Are the individual baby brains aligned to a common surface? Or is it an average of each of the 34 rois, averaged across the babies and projected on a sample surface?

6. I really liked the functional connectivity analysis, however, I am unclear what the non-homotopic means in this case. Can the authors also explicate what negative connectivity would mean and if it is significant what it would imply? It is also unclear how a correlation of r=.09 in Figure 4b would be significant and passes a multiple comparison threshold. Please clarify.

7. In "Comparison between structural and functional properties in ventral visual cortex of term- and preterm-born infants" analysis: The juxtaposition of these two statements is very confusing to me: "For the structural measurements, 56 of 68 ROIs showed significant lower CT in the term-born than preterm-born infants (ts = -4.26 to -2.71; FDR q < 0.05; Figure 6b). Particularly, the term born infants showed significantly higher CT in the area V1 (t = -2.55, p < 0.02) but not VOTC (t = -1.24, p > 0.2; Figure 6f)." I would rather see a mean CT (mean/std) across the rois in numbers and how it varies between preterm and term-born infants.

8. Relatedly in figure 6b-c, it is difficult to know the significant ts. Please change the non-significant ts to a different color or grays.

9. The analysis titled "Comparison between structural and functional properties in ventral visual cortex of term- and preterm-born infants" is a bit confusing to me. I am not clear on how we learn anything about the structural-functional relationship. It would be more convincing to see some comparative analysis between the structural and functional measures and how this varies by age, and in pre and full-term infants. The current analysis is not doing justice to the title of this analysis.

10. Your results of more myelination in the V1 than higher-order areas are consistent with the hierarchical findings of a prior paper on infants in their first six months of life (see Natu, et al., 2021, Nature Communications Biology). What is interesting to see is that neural scaffolding starts to develop early on during prenatal life. However, it is also interesting that you found that the cortical myelination in the ventral cortex was not directly influenced by postnatal time and I think these points can be discussed in the discussion.

11. Finally, I am not entirely clear if the term postnatal "experience" used throughout the manuscript, is appropriate in this case, perhaps the authors can tone it down the postnatal time, as we don't have any behavioral measure of visual experience per se, like contrast properties, acuity, etc.

Reviewer #2 (Recommendations for the authors):

Most of my recommendations are integrated into the public comments. That said, I wish to emphasize how important I think this question is and the impressive amount of work the authors have already done.

A general style comment is that I had trouble following the thread in the results. I think that each analysis should be framed more clearly in terms of the question it is answering. It was often difficult to understand how an analysis being described would contribute to the overall claims of the paper and what alternative hypotheses were possible for the analysis. This led to what felt like a list of results, rather than a coherent narrative.

Recommended analyses:

Regarding point 3: An alternative to this analysis I think the authors ought to consider is to control PT (i.e., find preterm and term groups that have an equivalent postnatal age) and see how variability in GA contributes to the metrics they observe. This reverses the logic of the analysis the authors did, but does so in a way I think is tightly controlled.

Regarding point 9: an alternative test that compensates for some of these constraints is to get the homotopic correlation for a region and then subtract it from the average of the non-homotopic correlations for that region (could even just use the distal pairs for this). This difference can then be compared across individuals for each region.

Regarding point 10: the authors should consider using the Wang et al., (2015) atlas, since that more carefully divides visual cortex.

Reviewer #3 (Recommendations for the authors):

– The language at times gets confusing referring to things such as PMA, GA, PA, etc. So I would simplify and clarify the language and say "gestational time" or "gestational age" when referring to how long a neonate spent gestating and "postnatal time" when referring to how much experience it has had after birth. Gestational time and postnatal time are not that long character-wise and will go a long way in improving the readability of the manuscript. I found myself having to go back and look up acronym meanings several times.

– In figure 2, the outline is hard to see in some panels. For example, in Figure 2a, it at first looks like only that central fusiform ROI is the only ROI with a significant value, but I didn't realize until later that it's actually all of the other ROIs that seem to be outlined and the fusiform ROI is not significant. I would say that if an ROI is not significant, I would just black it out, that way if an ROI is colored it is because that value is significant. This will be more straightforward than the current outlines.

– Supp Figure 4 seems important why isn't it a main figure or part of main figure 3? 3 could be condensed or pruned to make room for it. Also on the topic of supplementary figure 4, from what ROI was this taken? All of the visual HCP-MMP ROIs (like V1 + high-level regions) or was it one particular region? If incorporated into Figure 3, showing the data for cortical myelination like parts D and E would be great.

– Figure 4d, would it be worthwhile to make this figure comparing infants with longer PMA times since it was shown that functional connectivity network properties change with development? It might be nice to see how and if the network in Figure 4D changes (do some regions get closer together, etc).

– Typo in figure titles in Figure 5.

– Page 15 Line 3 states that term-born infants have a lower cortical thickness, but then on line 5 it says that term-born infants have a higher cortical thickness in V1. I think this is a typo, Figure 6F would suggest term-born have thinner V1.

– On the topic of Figure 6, it is being shown that the cortex is thinner in V1 in term-born infants who have comparatively less postnatal time. It is also shown that these term-born infants have more myelin. Typically, myelin increases with development, and the cortical thinning that occurs later in childhood is thought to be a result of increasing myelination at the gray-white matter border (see Natu et al. 2019 PNAS). Is it possible some of this myelin difference could result from the fact that in term-born infants the cortex is thinner (1.4mm in V1) and thus some voxels could partial-volume white matter voxels more easily? I know that white matter in neonates appears darker than the cortex in a T1-weighted image, but just want to bring up this potential point that a thinner cortex, given constant-sized voxels, is more likely to be biased by white matter signal from voxels sampling deep cortical layers.

– It might be nice to have a summary table showing, for each measure you compared against gestational time and prenatal time, whether the gestational or prenatal time was a significant modulator of that measure. Since one of the big conclusions of this paper is that not all structural and functional aspects are affected by gestation or postnatal time the same way, having a quick summary of those findings would be helpful.

– In the last Results section "Comparison between structural and functional properties…", I was expecting the authors to more formally compare structural and functional development. I was expecting them to do a regression with functional connectivity and structural metrics, to ask for example if a region's homotopic connectivity is correlated with CT or CM and which aspect of time (gestation or postnatal) is most important. You could even model gestational time and postnatal time separately. I say this because the data currently in Figure 6 while certainly useful, are summaries of data already being shown in Figures 3-5. One way to expand this would be to run the model I just mentioned. It would also be helpful in 6B-D to label what red or blue means (including labeled colors in B and C would help too). I know it's in the legend but it doesn't hurt to directly label to aid in making it quickly accessible.

– Page 16 line 10, what does developmental maturity mean? Please use gestation time or postnatal time or PMA to clarify what you mean here.

https://doi.org/10.7554/eLife.78733.sa1

Author response

Essential revisions:

Major Concerns:

1. One important concern brought up by all three reviewers is the confusion with the terminologies GA, PMA, and post-natal time (PT), for preemies/full-term infants. Please define and simplify terms as suggested by the reviewers to ameliorate the readability and justify why the authors chose to use each of the terms at different times in the analyses. Overall, the reviewers suggested using GA and PT and replacing the analyses with GA to simplify the findings. Please see individual reviewers' comments to simplify the language and avoid confusion across the manuscript.

We are sorry for the confusion caused by the terminologies. We have replaced all the analyses involving PMA with GA, and thus only GA and PT were used in the manuscripts to simplify the terms and ameliorate the readability, except for Figure 1 which showed the general development trajectories of structural or functional properties in the visual cortex. Figures 2-6 and the corresponding results and discussions are all updated.

2. To make the results and the flow of the paper more cohesive, we recommend the authors frame and justify each analysis with a clear question that the analysis will answer or the hypothesis that the authors posit as it is often difficult to understand how an analysis being described would contribute to the overall claims of the paper.

Thanks for the suggestions, we have clarified descriptions of the hypothesis and questions for each of the analyses, and also added a summary of these questions and manuscript organization at the end of the introduction:

“In the present study, we first characterized the general development of structural morphology (CT) and microstructure (CM) in the ventral cortex and investigate the contributions of prenatal and postnatal time on the structural development. Then we used r-fMRI to characterize the innate organization of the ventral cortex as early as one day of life and investigated the effect of PT on functional networks of V1 and VOTC areas. Furthermore, we carried out a mediation analysis to investigate the relationship between functional and structural development in area V1. Finally, we evaluated the structural and functional differences of visual cortex between the term- and preterm-born babies, as the latter group had immature cortical development at birth but longer PT compared to the former group with equivalent PMA”.

The rationale for individual results was added before each of the analyses:

1) Before the section “General development of cortical thickness and cortical myelination in human infants” (figure 1), we added “To show the general developmental trajectories of CT and CM in the visual cortex during the human neonatal period” (Page 5 Line 19-20);

2) Before the section of “Contribution of postnatal time and gestational age on the development of cortical thickness and cortical myelination” (figure 2), we added “Similar developmental trajectories of CT and CM in the ventral cortex across PMA were observed in the above results. It would be interesting to know whether gestational age (GA, the time from fertilization to birth) and postnatal time (PT, the time from birth to MRI scan) would have different influences on their development because the PT and GA reflected two very factors (e.g. innate growth versus postnatal experience)” (Page 6 Line 19-21 and Page 7 Line 1-2).

3) Before the section of “Development of functional connections in the ventral visual cortex” (figure 4), we added

“Beyond structural development of the neonatal cortex, we further asked how the functional connectivity between the different visual sub-regions changes during early development” (Page 9 Line 11-13) and

“Based on the baseline functional connections in term-born neonates, we were intrigued to known how those functional properties in the ventral visual cortex develop during neonatal stage of human infants? If so, whether PT and GA had different influences on this process?” (Page 10 Line 11-14).

3. Relatedly, the reviewers also brought up important points in several places whereby the claims made in the paper are not supported by the analyses conducted under this claim (e.g.., in cortical myelination and overall functional connectivity analyses in Figure 2,4, and 5 as well as the final results). The statistical tests and analysis conducted need to support the conclusions and claims. We recommend that the authors rephrase or tone down the conclusions made in this analysis or reassess their analyses to support the claims. A similar concern is raised by reviewers 1 and 3 in the final section "Comparison between structural and functional properties…", as the reader is left expecting a formal comparison between structural and functional development. Please consider an analysis as suggested by reviewer 3 for e.g., a regression with functional connectivity and structural metrics, to ask for example if a region's homotopic connectivity is correlated with CT or CM and which aspect of time (gestation or postnatal) is most important.

We have added a formal analysis of the structural-functional relationship to examine if a region's homotopic connectivity is correlated with CT or CM and which aspect of time (gestation or postnatal). Please see details in response to R1Q1 below.

4. Given the authors used an adult atlas to draw regions on the infants' brains, a sanity check of how well the Glasser atlas regions align on a neonatal brain at e.g., 37 weeks and compare that to 42 weeks old and adult brain and running some statistics on how these cortical-based alignments from the adult brain to the neonatal brain align, for a few sample subjects is recommended.

We have newly added analysis using the Glasser’s atlas and compared the area V1 transformed from adult space and manually delineated V1 in the neonatal space. Please see the response to the R1Q5.

5. A similar sanity check is also essential to check the white-gray matter boundary segmentation in some sample infant data as the authors use cortical thickness as one of their measures and CT is measured as the distance between the pial surface and gray-white matter boundary and the cortical ribbon of infants is thin at birth, there may be a possibility that partial-volume effects could be more prevalent in less-developed infants and impact myelin metrics.

We have added a sanity check on the segmentation accuracy of white-gray boundary in Figure 1—figure supplement 3. See the response to R1Q5.

6. The reviewers also asked if the authors used Fisher Z to transform their correlations? This is important to clarify as a significant portion of the results in the paper are correlations and it is unclear from the methods whether the authors are transforming their correlation values to make that use appropriate.

We are sorry for the confusion. All the statistical analyses involving correlation coefficients were Fisher-Z transformed. We have clarified this point in the revised manuscript.

7. Since one of the big conclusions of this paper is that not all structural and functional aspects are affected by gestation or postnatal time the same way, having a quick summary of those findings would be helpful. The reviewers suggest a summary table showing the morphological and myelin measures against gestational time and prenatal time, whether the gestational or prenatal time was a significant modulator of that measure.

Thank you for this suggestion. A summary table (Table 1) has been added to the manuscript.

8. Reviewer 2 brings up an important point about the homotopy analysis that needs clarification and perhaps rethinking of the statistical tests (e.g., comparing 34 homotopic pairs with the hundreds of non-homotopic pairs) that must be addressed in the revision. Please also consider the recommended alternative test by reviewer 2 that would compensate for some of the constraints to measure the homotopic correlation for a region.

We have modified the homotopic analysis following the suggestions of reviewer 2, by defending bilateral connections in to three categories of homotopic connections, adjacent connections and distal connections. Please see the specific responses to R2Q9 below.

Reviewer #1 (Recommendations for the authors):

While I find this work theoretically well-motivated and the use of the large dHCP dataset very exciting, there are some major concerns that must be addressed before this study is ready for publication:

1. There was a lot of confusion for me about the terminology between GA, PMA, and post-natal time, for preemies/full-term infants. To simplify the matter, I would first define these ages for the reader and also choose to use a single term, preferably GA as postmenstrual age would be GA plus the chronological age. This would make it easy to follow the flow of the analysis, especially in the section: "Contribution of postnatal time on the development of cortical thickness and cortical myelination". I also feel the same about the analysis in figure 3, it would make it clearer to use GA and postnatal time, as PMA already included both and it's difficult to see the relationship of the GA on CT and CM separately.

We agree it will be clearer to use GA and postnatal time instead of PMA and have revised the manuscript throughoutly. Please see our response to the 1st major concern in the Essential Revisions (for the authors) section above.

2. Relatedly, I really like Supplementary Figure 4c and 4f, and they may be your point on the postnatal age and GA influence on CT and CM really clear and they can be moved as one of the main figures in the paper.

Thanks for the Reviewer’s suggestion. We have updated the results relating to Supp Figure S4 following the suggestion of Reviewer 2 using a linear mixed-effect model instead of the present ANOVA analysis to keep the continuity of the data (Page 7 Line 15-19):

“We applied a linear mixed-effect model to test whether the CT (or CM) of the whole ventral cortex were differently influenced by the GA vs. PT, and found that the GA had a significantly stronger effect on the CM than PT (interaction between GA and PT, p < 0.05) but no significant difference between GA and PT effect was found on the development of CT (p > 0.6)”.

Therefore, the previous Supplementary Figure 4 no longer exists, but the related information is described in text.

3. It would be good to sanity check how the adult Glasser maps align on a neonatal brain at e.g., 37 weeks and compare that to 42 weeks old and adult and run some statistics on how these cortical-based alignments from the adult brain to the neonatal brain align, for a few sample subjects. Another sanity check that is necessary especially with the cortical thickness measure is to check the segmentations in the data. It would be useful to provide a montage of a few sample subjects showing the segmentation of the gray-white boundary as the contrast properties in infants is hard to decipher.

Thank you for the suggestions. We agree it is necessary to provide a sanity check of the registration accuracy.

Regarding the registration between adult Glasser’s maps and neonatal brain, we didn’t register the adult maps into individual spaces of the neonates since this would need multiple steps of registrations, resulting in inaccuracies and the additional burden of checking the registration of each subject. Instead, we first registered all neonatal surfaces into a neonate-specific template (40-week) using the transformation matrices provided by the dHCP pipeline based on the multimodal surface matching method (MSM, Makropoulos et al., 2018; Robinson et al., 2018). Then we performed a similar MSM registration to transform Glasser’s maps into the 40-week template. We visually checked the registration results for some of the typical sulci and gyri in Figure 1—figure supplement 1. Although the best way is to use a neonate-specific parcellation for the present analysis, there is no such fine cortical parcellation of human neonates yet. In addition, we used a definable area—the V1 area as an example to quantitatively evaluate the quality of the registration. Specifically, we manually delineated the area V1 on the gradient map (Figure 1—figure supplement 4a) of the averaged CM map across all neonates and calculate the dice coefficient between the transformed V1 from Glasser’s atlas and the manually delineated V1. The result showed a dice score of 0.83 and 0.80 for right and left hemispheres, respectively. As shown in Figure 1—figure supplement 4b, the location of the registered V1 and manually delineated V1 area was consistent, although the latter was slightly bigger. Identifying other visual areas in Glasser’s map would need task functional data that were not available in the present study. These quality checks have been added in the Methods section (Page 23 Line 23-29 and Page 24 Line 1-5).

Regarding the gray-white matter segmentation, we used the dHCP pipeline in this work. The segmentation procedure was included in the dHCP pipeline, and developers visually checked the quality of segmentation in their study, in which only 5% of the subjects had poor segmentation based on 160 random images from 37-44 weeks of PMA (Makropoulos et al., 2018). Here we provided some examples of the gray-white boundaries from the subjects scanned at different ages in Figure 1—figure supplement 3 (Page 22 Line 19).

4. In the first analysis are there any hemispheric differences in the relationship between CT and CM and age?

We applied a linear mixed-effect model to estimate the hemispheric difference in the relationship between structural measurements and age with the hemisphere, age, and the interaction between them as a fixed effect. The intercept and coefficient of the hemisphere were the random effects. For the averaged measurements across the whole ventral mask, the main effect of age was significant in both CT and CM (p < 0.05), but the main effect of the hemisphere and the interaction between age and hemisphere were not (p > 0.3).

5. Can the authors describe how the average cortical thickness and myelination maps in Figures1 b, c are generated? Are the individual baby brains aligned to a common surface? Or is it an average of each of the 34 rois, averaged across the babies and projected on a sample surface?

The average cortical thickness and myelination maps are generated by aligning individual baby brains to a common surface and then average. We have clarified this procedure in the Methods section (Page 26 Line 1-6):

“To generate the average cortical measurements (CT and CM) at the group level, we first registered the metrics of every subject onto a common 40-week template surface using the registration files provided by dHCP pipeline, and then averaged them in a vertex-wise manner across subjects within a specific week (e.g. 38-week map included neonates whose PMA ranged from 37.5 to 38.5 weeks) “.

6. I really liked the functional connectivity analysis, however, I am unclear what the non-homotopic means in this case. Can the authors also explicate what negative connectivity would mean and if it is significant what it would imply? It is also unclear how a correlation of r=.09 in Figure 4b would be significant and passes a multiple comparison threshold. Please clarify.

The non-homotopic indicated the pairs between two non-paired areas in two hemispheres, e.g., V1 in the left hemisphere and V2 in the right. In the present study, only 34 paired areas are homotopic (diagonal of the connectivity matrix) and all other connections were non-homotopic. Now we have changed them into three groups following the suggestion by Reviewer 2: “namely the homotopic connection (the connection between two paired areas in two hemispheres. e.g. right and left V1), adjacent connection (e.g., right V1 and left V2 as well as left V1 and right V2 since V1 and V2 are adjacent) and distant connections (two areas that were not the paired or adjacent).” We have added this description in the Methods section (Page 26 Line 24-29).

It’s always been controversial whether the negative functional connectivity (also called anti-correlation) is valid neurophysiologically or just analytic artifacts (Chai et al., 2012; Murphy et al., 2009). Some researchers might take the anti-correlation as an inhibitory interaction in the specific circuit if it was not driven by the analytic procedure (Keller et al., 2015; Meskaldji et al., 2016). In the revised paper, we excluded the negative connections to avoid such uncertainty.

Sorry for the mistake about the “r = 0.09 in Figure 4b”. We mistakenly calculated the p-value based on Pearson correlation using the number of time points as the degree of freedom. Now we updated the statistical analysis using an independent-sample T-test to test whether the homotopic correlation was significantly greater than zero across subjects, and the results showed that the homotopic connections in all ROIs of ventral cortex were significant (mean r = 0.13– 0.43, t > 12.87, p < 10-9; Figure 4a-b). See corrected parts in the Methods and Results (Page 9 Line 17-21).

7. In "Comparison between structural and functional properties in ventral visual cortex of term- and preterm-born infants" analysis: The juxtaposition of these two statements is very confusing to me: "For the structural measurements, 56 of 68 ROIs showed significant lower CT in the term-born than preterm-born infants (ts = -4.26 to -2.71; FDR q < 0.05; Figure 6b). Particularly, the term born infants showed significantly higher CT in the area V1 (t = -2.55, p < 0.02) but not VOTC (t = -1.24, p > 0.2; Figure 6f)." I would rather see a mean CT (mean/std) across the rois in numbers and how it varies between preterm and term-born infants.

Thanks for the suggestions. We added bar figures to show the mean CT and CM across all 68 ROIs bilaterally between term and preterm-born infants and revised the descriptions as “For the structural measurements, the mean CT across all ROIs was significantly lower in the term- (1.39 ± 0.05) than preterm-born infants (1.35 ± 0.05; t = -3.16, p < 0.01) and 10 of 68 ROIs showed significantly lower CT in the term-born than preterm-born infants (Figure 7—figure supplement 1). On the contrary, the mean CM across all ROIs was higher in the term-born (1.22 ± 0.10) than preterm-born (1.09 ± 0.08; t = 7.11, p < 10-9) infants and all ROIs were significantly higher in the term- than preterm-born groups.” in the Results section (Page 15 Line 3-9). Meanwhile, we found that the previous statement “56 of 68 ROIs showed significantly lower CT in the term-born than preterm-born infants” was problematic. The correct result is that 56 of 68 ROIs showed lower CT in the term-born than preterm-born infants and 10 of them were significant. We have corrected the statement in the above changes.

8. Relatedly in figure 6b-c, it is difficult to know the significant ts. Please change the non-significant ts to a different color or grays.

We have revised Figure 6b-c accordingly.

9. The analysis titled "Comparison between structural and functional properties in ventral visual cortex of term- and preterm-born infants" is a bit confusing to me. I am not clear on how we learn anything about the structural-functional relationship. It would be more convincing to see some comparative analysis between the structural and functional measures and how this varies by age, and in pre and full-term infants. The current analysis is not doing justice to the title of this analysis.

Please see the response to R1Q1.

10. Your results of more myelination in the V1 than higher-order areas are consistent with the hierarchical findings of a prior paper on infants in their first six months of life (see Natu, et al., 2021, Nature Communications Biology). What is interesting to see is that neural scaffolding starts to develop early on during prenatal life. However, it is also interesting that you found that the cortical myelination in the ventral cortex was not directly influenced by postnatal time and I think these points can be discussed in the discussion.

Thanks for the advice, we added the following paragraph in the Discussion section (Page 18 Line 17-23):

“However, we found a significant correlation between CM in the ventral cortex and GA but not PT in the neonatal period, suggesting that early cortical myelination was primarily determined by the endogenic growth in the uterine environment. The hierarchical pattern of CM along the ventral visual cortex agreed with a previous study on infants at around six months of age (Natu et al., 2021). Our results further suggested that this hierarchical pattern occurs as early as the neonatal period.”

11. Finally, I am not entirely clear if the term postnatal "experience" used throughout the manuscript, is appropriate in this case, perhaps the authors can tone it down the postnatal time, as we don't have any behavioral measure of visual experience per se, like contrast properties, acuity, etc.

We have toned it down to the postnatal time or postnatal environment in the revised article.

Reviewer #2 (Recommendations for the authors):

Most of my recommendations are integrated into the public comments. That said, I wish to emphasize how important I think this question is and the impressive amount of work the authors have already done.

A general style comment is that I had trouble following the thread in the results. I think that each analysis should be framed more clearly in terms of the question it is answering. It was often difficult to understand how an analysis being described would contribute to the overall claims of the paper and what alternative hypotheses were possible for the analysis. This led to what felt like a list of results, rather than a coherent narrative.

Thanks for the suggestions, we have added clearer descriptions of the hypothesis and questions for each of the analyses to put them in a coherent story in the revised manuscript. Please see the Response to the 2nd major concern in the Essential Revisions (for the authors) section above.

Recommended analyses:

Regarding point 3: An alternative to this analysis I think the authors ought to consider is to control PT (i.e., find preterm and term groups that have an equivalent postnatal age) and see how variability in GA contributes to the metrics they observe. This reverses the logic of the analysis the authors did, but does so in a way I think is tightly controlled.

We agree with this view, but the present data might not be suitable for this analysis. The range of the PT in the term-born infants was from 0 to 6.72 weeks (mean value: 1.12 ± 1.25) but PT in the preterm-born infants was from 0.28 to 19.72 weeks (mean value: 8.91± 4.42). As shown in Author response image 1, the distribution of PT was also very different in the term and preterm groups. Therefore, we simply controlled for GA or PT in the term-born infants to see the effect of the other one (Page 7 Line 20-28, and Figure 2 e-h).

Author response image 1
The distribution of neonates with different PT in the term and preterm infants.

Regarding point 9: an alternative test that compensates for some of these constraints is to get the homotopic correlation for a region and then subtract it from the average of the non-homotopic correlations for that region (could even just use the distal pairs for this). This difference can then be compared across individuals for each region.

We have performed a non-parameter statistical analysis (Wilcoxon signed-rank) to statistically compare homotopic connections and other types of connections. Please see the Response to R2Q9.

Regarding point 10: the authors should consider using the Wang et al., (2015) atlas, since that more carefully divides visual cortex.

The difference between Glasser’s atlas and Wang’s atlas is that Wang’s atlas is focused on the human visual cortex and provided 25 areas per hemisphere, including 8 early visual cortex (e.g. V1-3), 4 lateral occipital areas (e.g. LO1, LO2), 5 ventral visual cortex (e.g. hV4, PHC), 7 parietal regions (e.g. IPS) and 1 fontal area (e.g. hFEF). While the Glasser’s atlas contained 180 areas per hemisphere for the whole brain cortex with 17 ROIs in the ventral visual cortex p. Wang’s atlas provided a more detailed parcellation than Glasser’s atlas in the early visual cortex (e.g. the area V1 was divided into V1v and V1d in the Wang’s atlas). Therefore, we tried to use Wang et al., (2015) atlas to reanalysis the data.

We registered Wang’s atlas onto 40-week neonatal template by the same method applied to Glasser’s atlas (see Author response image 2), and then calculated the homotopic connection for the 17 ROIs within the visual cortex in the Wang’s atlas (we excluded areas in the parietal and frontal lobe). The resulting connections ranged from 0.001 to 0.55, and the mean homotopic connection was 0.23 (SD = 0.17). We found that the correlation between bilateral V1 was improved while some small ROIs showed very low correlations. Those low value might be due to the registration inaccuracy that these refined areas were misregistered into non-homologous positions between hemispheres, and therefore, we think Wang’s atlas with very small ROIs may not be appropriate for the task here.

Author response image 2
Mapping of Wang et al.

(2015) atlas on the 40-week neonatal template.

Reviewer #3 (Recommendations for the authors):

– The language at times gets confusing referring to things such as PMA, GA, PA, etc. So I would simplify and clarify the language and say "gestational time" or "gestational age" when referring to how long a neonate spent gestating and "postnatal time" when referring to how much experience it has had after birth. Gestational time and postnatal time are not that long character-wise and will go a long way in improving the readability of the manuscript. I found myself having to go back and look up acronym meanings several times.

We have defined GA as the "gestational age" referring to the period a neonate spent gestating and PT as the "postnatal time" referring to how much experience it has had after birth. Since we have throughoutly revised the analyses using only GA and PT as the developmental factors, we think it would make the manuscript easier to read now.

– In figure 2, the outline is hard to see in some panels. For example, in Figure 2a, it at first looks like only that central fusiform ROI is the only ROI with a significant value, but I didn't realize until later that it's actually all of the other ROIs that seem to be outlined and the fusiform ROI is not significant. I would say that if an ROI is not significant, I would just black it out, that way if an ROI is colored it is because that value is significant. This will be more straightforward than the current outlines.

We have updated figure 2 in the new manuscript to only show the significant ROIs, and also incorporated comments from all reviewers.

– Supp Figure 4 seems important why isn't it a main figure or part of main figure 3? 3 could be condensed or pruned to make room for it. Also on the topic of supplementary figure 4, from what ROI was this taken? All of the visual HCP-MMP ROIs (like V1 + high-level regions) or was it one particular region? If incorporated into Figure 3, showing the data for cortical myelination like parts D and E would be great.

Thanks for the Reviewer’s suggestion. We have updated the results relating to Supp Figure S4 following the suggestion of Reviewer 2 using a linear mixed-effect model rather than the present ANOVA analysis to keep the continuity of the data (Page 7 Line 15-19):

“We applied a linear mixed-effect model to test whether the CT (or CM) of the whole ventral cortex were differently influenced by the GA vs. PT, and found that the GA had a significantly stronger effect on the CM than PT (interaction between GA and PT, p < 0.05) but no significant difference between GA and PT effect was found on the development of CT (p > 0.6).”

Therefore, the previous Supplementary Figure 4 no longer exists.

– Figure 4d, would it be worthwhile to make this figure comparing infants with longer PMA times since it was shown that functional connectivity network properties change with development? It might be nice to see how and if the network in Figure 4D changes (do some regions get closer together, etc).

We add a new analysis to directly evaluate the developmental change of the functional connections in the visual cortex with PMA (Page 11 Line 15-19):

“The correlations between ipsilateral connections and PMA showed both positive and negative results in different regions of the ventral cortex. Particularly, the early visual cortex (e.g. V1 and V2) showed decreasing connection to other visual areas across PMA while part areas in the VOTC (e.g. fusiform and parahippocampus) showed increasing tendency (Figure 4—figure supplement 4).”

– Typo in figure titles in Figure 5.

We have revised accordingly.

– Page 15 Line 3 states that term-born infants have a lower cortical thickness, but then on line 5 it says that term-born infants have a higher cortical thickness in V1. I think this is a typo, Figure 6F would suggest term-born have thinner V1.

We have corrected the typo. The term-born infants had a lower cortical thickness in V1 than preterm-born infants.

– On the topic of Figure 6, it is being shown that the cortex is thinner in V1 in term-born infants who have comparatively less postnatal time. It is also shown that these term-born infants have more myelin. Typically, myelin increases with development, and the cortical thinning that occurs later in childhood is thought to be a result of increasing myelination at the gray-white matter border (see Natu et al. 2019 PNAS). Is it possible some of this myelin difference could result from the fact that in term-born infants the cortex is thinner (1.4mm in V1) and thus some voxels could partial-volume white matter voxels more easily? I know that white matter in neonates appears darker than the cortex in a T1-weighted image, but just want to bring up this potential point that a thinner cortex, given constant-sized voxels, is more likely to be biased by white matter signal from voxels sampling deep cortical layers.

We agree that there is a possibility that thinning of the cortex and change of the cortical contrast due to myelination could contribute to our observation. Although we have done our best to ensure the segmentation accuracy (see the response to R1Q5), there’s no way to exclude this effect. We added a paragraph in the limitation section to describe the possible bias driven by the thinner cortex in the term-born infants compared to preterm-born infants (Page 20 Line 20-24):

“And also the term-born infants, who had thinner cortex compared to preterm-born infants, were more likely to be affected by the partial-volume effect, which may contribute to the myelination difference observed between groups.”

– It might be nice to have a summary table showing, for each measure you compared against gestational time and prenatal time, whether the gestational or prenatal time was a significant modulator of that measure. Since one of the big conclusions of this paper is that not all structural and functional aspects are affected by gestation or postnatal time the same way, having a quick summary of those findings would be helpful.

We appreciate the reviewer’s suggestion, and we have provided a summary table.

– In the last Results section "Comparison between structural and functional properties…", I was expecting the authors to more formally compare structural and functional development. I was expecting them to do a regression with functional connectivity and structural metrics, to ask for example if a region's homotopic connectivity is correlated with CT or CM and which aspect of time (gestation or postnatal) is most important. You could even model gestational time and postnatal time separately. I say this because the data currently in Figure 6 while certainly useful, are summaries of data already being shown in Figures 3-5. One way to expand this would be to run the model I just mentioned. It would also be helpful in 6B-D to label what red or blue means (including labeled colors in B and C would help too). I know it's in the legend but it doesn't hurt to directly label to aid in making it quickly accessible.

Following the suggestions, we have added a formal analysis to investigate the relationship between structural and functional development in area V1. Please see the Response to the 1st concern of Reviewer 1 (public review) for more details.

– Page 16 line 10, what does developmental maturity mean? Please use gestation time or postnatal time or PMA to clarify what you mean here.

We replaced developmental maturity with GA in the manuscript to make it clear.

References:

Chai XJ, Castañán AN, Öngür D, Whitfield-Gabrieli S. 2012. Anticorrelations in resting state networks without global signal regression. Neuroimage 59:1420–1428. doi:10.1016/j.neuroimage.2011.08.048

Fjell AM, Grydeland H, Krogsrud SK, Amlien I, Rohani DA, Ferschmann L, Storsve AB, Tamnes CK, Sala-Llonch R, Due-Tønnessen P, Bjørnerud A, Sølsnes AE, Håberg AK, Skranes J, Bartsch H, Chen CH, Thompson WK, Panizzon MS, Kremen WS, Dale AM, Walhovd KB. 2015. Development and aging of cortical thickness correspond to genetic organization patterns. Proc Natl Acad Sci U S A 112:15462–15467. doi:10.1073/pnas.1508831112

Kamps FS, Hendrix CL, Brennan PA, Dilks DD. 2020. Connectivity at the origins of domain specificity in the cortical face and place networks. Proc Natl Acad Sci U S A 117:6163–6169. doi:10.1073/pnas.1911359117

Keller JB, Hedden T, Thompson TW, Anteraper SA, Gabrieli JDE, Whitfield-Gabrieli S. 2015. Resting-state anticorrelations between medial and lateral prefrontal cortex: Association with working memory, aging, and individual differences. Cortex 64:271–280. doi:10.1016/j.cortex.2014.12.001

Makropoulos A, Robinson EC, Schuh A, Wright R, Fitzgibbon S, Bozek J, Counsell SJ, Steinweg J, Vecchiato K, Passerat-Palmbach J, Lenz G, Mortari F, Tenev T, Duff EP, Bastiani M, Cordero-Grande L, Hughes E, Tusor N, Tournier JD, Hutter J, Price AN, Teixeira RPAG, Murgasova M, Victor S, Kelly C, Rutherford MA, Smith SM, Edwards AD, Hajnal J V., Jenkinson M, Rueckert D. 2018. The developing human connectome project: A minimal processing pipeline for neonatal cortical surface reconstruction. Neuroimage 173:88–112. doi:10.1016/j.neuroimage.2018.01.054

Meskaldji DE, Preti MG, Bolton TA, Montandon ML, Rodriguez C, Morgenthaler S, Giannakopoulos P, Haller S, Van De Ville D. 2016. Prediction of long-term memory scores in MCI based on resting-state fMRI. NeuroImage Clin 12:785–795. doi:10.1016/j.nicl.2016.10.004

Murphy K, Birn RM, Handwerker DA, Jones TB, Bandettini PA. 2009. The impact of global signal regression on resting state correlations: Are anti-correlated networks introduced? Neuroimage 44:893–905. doi:10.1016/j.neuroimage.2008.09.036

Natu VS, Rosenke M, Wu H, Querdasi FR, Kular H, Lopez-Alvarez N, Grotheer M, Berman S, Mezer AA, Grill-Spector K. 2021. Infants’ cortex undergoes microstructural growth coupled with myelination during development. Commun Biol 4:1–12. doi:10.1038/s42003-021-02706-w

Robinson EC, Garcia K, Glasser MF, Chen Z, Coalson TS, Makropoulos A, Bozek J, Wright R, Schuh A, Webster M, Hutter J, Price A, Cordero Grande L, Hughes E, Tusor N, Bayly P V., Van Essen DC, Smith SM, Edwards AD, Hajnal J, Jenkinson M, Glocker B, Rueckert D. 2018. Multimodal surface matching with higher-order smoothness constraints. Neuroimage 167:453–465. doi:10.1016/j.neuroimage.2017.10.037

Wang F, Lian C, Wu Z, Zhang H, Li T, Meng Y, Wang L, Lin W, Shen D, Li G. 2019. Developmental topography of cortical thickness during infancy. Proc Natl Acad Sci U S A 116:15855–15860. doi:10.1073/pnas.1821523116

https://doi.org/10.7554/eLife.78733.sa2

Article and author information

Author details

  1. Mingyang Li

    Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1639-9891
  2. Tingting Liu

    Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    Contribution
    Data curation, Software, Validation, Investigation, Methodology
    Competing interests
    No competing interests declared
  3. Xinyi Xu

    Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    Contribution
    Software, Validation, Visualization, Methodology
    Competing interests
    No competing interests declared
  4. Qingqing Wen

    Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    Contribution
    Visualization, Methodology
    Competing interests
    No competing interests declared
  5. Zhiyong Zhao

    Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    Contribution
    Software, Methodology
    Competing interests
    No competing interests declared
  6. Xixi Dang

    Department of Psychology, Zhejiang Sci-Tech University, Hangzhou, China
    Contribution
    Conceptualization, Investigation
    Competing interests
    No competing interests declared
  7. Yi Zhang

    Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    Contribution
    Resources, Data curation
    Competing interests
    No competing interests declared
  8. Dan Wu

    1. Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
    2. Children's Hospital School of Medicine, Zhejiang University, Hangzhou, China
    3. Binjiang Institute of Zhejiang University, Hangzhou, China
    Contribution
    Conceptualization, Resources, Data curation, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Project administration, Writing - review and editing
    For correspondence
    danwu.bme@zju.edu.cn
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9303-5821

Funding

The Ministry of Science and Technology of the People's Republic of China (2018YFE0114600)

  • Dan Wu

National Natural Science Foundation of China (81971606)

  • Dan Wu

Science and Technology Department of Zhejiang Province (202006140)

  • Dan Wu

The Ministry of Science and Technology of the People's Republic of China (2021ZD0200202)

  • Dan Wu

National Natural Science Foundation of China (82122032)

  • Dan Wu

Science and Technology Department of Zhejiang Province (2022C03057)

  • Dan Wu

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by the Ministry of Science and Technology of the People’s Republic of China (2018YFE0114600, 2021ZD0200202), National Natural Science Foundation of China (81971606, 82122032), and Science and Technology Department of Zhejiang Province (202006140, 2022C03057). Data were provided by the developing Human Connectome Project, KCL-Imperial-Oxford Consortium funded by the European Research Council under the European Union Seventh Framework Programme (FP/2007–2013)/ERC Grant Agreement no. 319456. We are grateful to the families who generously supported this trial.

Ethics

The dHCP study was approved by the UK Health Research Authority (14/LO/1169) and written consent was obtained from the parents or legal guardians of the study subjects.

Senior Editor

  1. Joshua I Gold, University of Pennsylvania, United States

Reviewing Editor

  1. Vaidehi Natu, Stanford University, United States

Reviewer

  1. Cameron Ellis, Haskins Laboratories, United States

Publication history

  1. Received: March 17, 2022
  2. Preprint posted: March 18, 2022 (view preprint)
  3. Accepted: October 31, 2022
  4. Version of Record published: November 18, 2022 (version 1)

Copyright

© 2022, Li et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 152
    Page views
  • 30
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Mingyang Li
  2. Tingting Liu
  3. Xinyi Xu
  4. Qingqing Wen
  5. Zhiyong Zhao
  6. Xixi Dang
  7. Yi Zhang
  8. Dan Wu
(2022)
Development of visual cortex in human neonates is selectively modified by postnatal experience
eLife 11:e78733.
https://doi.org/10.7554/eLife.78733
  1. Further reading

Further reading

    1. Neuroscience
    Andrew P Davison, Shailesh Appukuttan
    Insight

    Artificial neural networks could pave the way for efficiently simulating large-scale models of neuronal networks in the nervous system.

    1. Neuroscience
    Jonathan Nicholas, Nathaniel D Daw, Daphna Shohamy
    Research Article

    A key question in decision making is how humans arbitrate between competing learning and memory systems to maximize reward. We address this question by probing the balance between the effects, on choice, of incremental trial-and-error learning versus episodic memories of individual events. Although a rich literature has studied incremental learning in isolation, the role of episodic memory in decision making has only recently drawn focus, and little research disentangles their separate contributions. We hypothesized that the brain arbitrates rationally between these two systems, relying on each in circumstances to which it is most suited, as indicated by uncertainty. We tested this hypothesis by directly contrasting contributions of episodic and incremental influence to decisions, while manipulating the relative uncertainty of incremental learning using a well-established manipulation of reward volatility. Across two large, independent samples of young adults, participants traded these influences off rationally, depending more on episodic information when incremental summaries were more uncertain. These results support the proposal that the brain optimizes the balance between different forms of learning and memory according to their relative uncertainties and elucidate the circumstances under which episodic memory informs decisions.