Skip to main content
NCBI home page
eLife logoLink to eLife
. 2023 Jan 24;12:e78397. doi: 10.7554/eLife.78397

Infant brain regional cerebral blood flow increases supporting emergence of the default-mode network

Qinlin Yu 1,2, Minhui Ouyang 1,2, John Detre 2,3, Huiying Kang 1,4, Di Hu 1,4, Bo Hong 5, Fang Fang 6, Yun Peng 4, Hao Huang 1,2,✉
Editors: Timothy E Behrens7, Timothy E Behrens8
PMCID: PMC9873253  PMID: 36693116

Abstract

Human infancy is characterized by most rapid regional cerebral blood flow (rCBF) increases across lifespan and emergence of a fundamental brain system default-mode network (DMN). However, how infant rCBF changes spatiotemporally across the brain and how the rCBF increase supports emergence of functional networks such as DMN remains unknown. Here, by acquiring cutting-edge multi-modal MRI including pseudo-continuous arterial-spin-labeled perfusion MRI and resting-state functional MRI of 48 infants cross-sectionally, we elucidated unprecedented 4D spatiotemporal infant rCBF framework and region-specific physiology–function coupling across infancy. We found that faster rCBF increases in the DMN than visual and sensorimotor networks. We also found strongly coupled increases of rCBF and network strength specifically in the DMN, suggesting faster local blood flow increase to meet extraneuronal metabolic demands in the DMN maturation. These results offer insights into the physiological mechanism of brain functional network emergence and have important implications in altered network maturation in brain disorders.

Research organism: Human

Introduction

The adult human brain receives 15–20% of cardiac output despite only representing 2% of body mass (Bouma and Muizelaar, 1990; Satterthwaite et al., 2014). Vast energy demand from the human brain starts from infancy, which is characterized by fastest energy expenditure increase across lifespan (Pontzer et al., 2021). Infancy is also the most dynamic phase of brain development across entire lifespan with fastest functional and structural brain development. For example, during infancy the brain size increases dramatically in parallel with rapid elaboration of new synapses, reaching 80–90% of lifetime maximum by age of year 2 (Knickmeyer et al., 2008; Ouyang et al., 2019a; Pfefferbaum et al., 1994). Structural and functional changes of infant brain are underlaid by rapid and precisely regulated (Huang et al., 2013; Silbereis et al., 2016) spatiotemporal cellular and molecular processes, including neurogenesis and neuronal migration (Rakic, 1995; Sidman and Rakic, 1973), synaptic formation (Huttenlocher and Dabholkar, 1997), dendritic arborization (Bystron et al., 2008; Ouyang et al., 2019b), axonal growth (Haynes et al., 2005; Innocenti and Price, 2005), and myelination (Miller et al., 2012; Yakovlev, 1967). These developmental processes demand rapidly increasing energy consumption of the brain. However, there have been few whole-brain mappings of heterogeneous infant brain regional cerebral blood flow (rCBF) changes across landmark infant ages from 0 to 24 months thus far, impeding understanding of energy expenditure across functional systems of early developing brain. As a result of differential neuronal growth across cortex, functional networks in the human brain develop differentially following the order from primary sensorimotor to higher-order cognitive systems (Cao et al., 2017a; Huang and Vasung, 2014; Sidman and Rakic, 1982; Tau and Peterson, 2010; Yu et al., 2016). The default-mode network (DMN) (Raichle et al., 2001) is widely recognized as a fundamental neurobiological system associated with cognitive processes that are directed toward the self and has important implication in typical and atypical brain development (Buckner et al., 2008). Unlike primary sensorimotor (SM) and visual (Vis) networks emerging relatively earlier around and before birth (Cao et al., 2017b; Doria et al., 2010; Smyser et al., 2010), emergence of the vital resting-state DMN is not well established until late infancy (Gao et al., 2009). Till date, it has been unclear how emergence of vital functional networks such as DMN is coupled with rCBF increase during infancy.

Regional brain metabolism, including glucose utilization and oxygen consumption, is closely coupled to regional CBF (rCBF) that delivers the glucose and oxygen needed to sustain metabolic needs (Raichle et al., 2001; Vaishnavi et al., 2010). Infant rCBF has been conventionally measured with positron emission tomography (PET) (Altman et al., 1988; Altman et al., 1993; Chugani and Phelps, 1986; Chugani et al., 1987) and single-photon emission computerized tomography (SPECT) (Chiron et al., 1992), which are not applicable to infants due to the associated exposure to radioactive tracers. By labeling the blood in internal carotid and vertebral arteries in neck and measuring downstream labeled arterial blood in brain, arterial-spin-labeled (ASL) (Alsop et al., 2015; Detre and Alsop, 1999) perfusion MRI provides a method for noninvasive quantifying rCBF without requiring radioactive tracers or exogenous contrast agents. Accordingly, ASL is especially suitable for rCBF measurements of infants (Ouyang et al., 2017; Lemaître et al., 2021; Wang et al., 2008) and children (Jain et al., 2012; Satterthwaite et al., 2014). Phase-contrast (PC) MRI, utilizing the phase shift proportional to velocity of the blood spins, has also been used to measure global CBF of the entire brain (Liu et al., 2019). Through integration of pseudo-continuous ASL (pCASL) and PC MRI, rCBF measured from pCASL can be calibrated by global CBF from PC MRI for more accurate infant brain rCBF measurement (Aslan et al., 2010; Ouyang et al., 2017). With rCBF closely related to regional cerebral metabolic rate of oxygen (CMRO2) and glucose (CMRGlu) at the resting state in human brain (Fox and Raichle, 1986; Gur et al., 2009; Paulson et al., 2010; Vaishnavi et al., 2010), rCBF could be used as a surrogate measure of local cerebral metabolic level for resting infant brains.

Early developing human brain functional networks can be reproducibly measured with resting-state fMRI (rs-fMRI). For example, a large scale of functional architecture at birth (Cao et al., 2017b; Doria et al., 2010; Fransson et al., 2007) has been revealed with rs-fMRI. Functional networks consist of densely linked hub regions to support efficient neuronal signaling and communication. These hub regions can be delineated with data-driven independent component analysis (ICA) of rs-fMRI data and serve as functional regions of interest (ROIs) for testing physiology–function relationship. Meeting metabolic demands in these hub ROIs is critical for functional network maturation. In fact, spatial correlation of rCBF to the functional connectivity (FC) in these functional ROIs was found in adult brains (Liang et al., 2013). Altered DMN plays a vital role in neurodevelopmental disorders such as autism (Doyle-Thomas et al., 2015; Lynch et al., 2013; Padmanabhan et al., 2017; Washington et al., 2014). Thus, understanding physiological underpinning of the DMN maturation offers invaluable insights into the mechanism of typical and atypical brain development.

We hypothesized heterogeneous rCBF maps at landmark infant ages and faster rCBF increase in brain regions of higher cognitive functions (namely DMN regions) during infancy than those of primary sensorimotor functions where functional networks emerge before or around birth (Cao et al., 2017a; Cao et al., 2017b; Doria et al., 2010; Fransson et al., 2007; Smyser et al., 2010; Peng et al., 2020). Furthermore, with rCBF as an indicator of local metabolic level of glucose and oxygen consumption, we hypothesized that strongly coupled rCBF and FC increase specifically in the DMN regions during infancy to meet extra metabolic demand of DMN maturation. In this study, we acquired multi-modal MRI, including both pCASL perfusion MRI, and rs-fMRI, of 48 infants aged 0–24 months to quantify rCBF and FC, respectively. RCBF at the voxel level and in functional network ROIs were measured to test the hypothesis of spatiotemporally differential rCBF increases during infancy. Maturation of FC in the DMN was delineated. Correlation of FC increase and rCBF increase in the DMN ROIs was tested and further confirmed with data-driven permutation analysis, the latter of which was to examine whether the coupling of rCBF and FC takes place only in the DMN during infancy.

Results

Emergence of the DMN during early brain development

Figure 1a shows the emergence of the DMN in typically developing brain from 0 to 24 months as measured using rs-fMRI with a posterior cingulate cortex (PCC, a vital hub of DMN network) seed region indicated by the black dash line. At around birth (0 months), the DMN is still immature with weak FC between PCC and other DMN regions, including medial prefrontal cortex (MPFC), inferior posterior lobule (IPL), and lateral temporal cortex (LTC) (Figure 1a). During infant brain development from 0 to 24 months, Figure 1a shows that the functional connectivity between MPFC, IPL, or ITC and PCC gradually strengthens. Figure 1b shows the FC–age correlation r value map. It can be appreciated from Figure 1b that across the cortical surface relatively higher r values are only located at the DMN regions (except the seed PCC).

Figure 1. Emergence of functional connectivity (FC) within the default-mode network (DMN) during infancy.

The maps of the DMN FC (PCC as a seed) at representative ages from 0 to 24 months are demonstrated in (a), and the map of correlation coefficient of FC (PCC as a seed) and age is demonstrated in (b). In (a), gradually emerging FC of other DMN regions (including MPFC, IPL, and LTC) to the PCC from 0 to 24 months can be appreciated. The PCC is delineated by the black dashed contour. In (b), stronger correlation between FC (PCC as a seed) and age is localized in DMN subregions IPL, ITL, and MPFC. Abbreviations of DMN subregions: IPL: inferior posterior lobule; LTC: lateral temporal cortex; MPFC: medial prefrontal cortex; PCC: posterior cingulate cortex.

Figure 1.

Figure 1—figure supplement 1. Identification of functional network regions of interest with resting-state fMRI of infants aged 12–24 months.

Figure 1—figure supplement 1.

(a) Functional connectivity obtained from correlation between blood oxygenation level-dependent (BOLD) fMRI time courses of two voxels. (b) Brain regions of the visual (Vis), sensorimotor (SM), and default-mode network (DMN) identified by z statistic maps from group independent component analysis. DMN subregions posterior cingulate cortex (PCC), medial prefrontal cortex (MPFC), inferior posterior lobule (IPL), and lateral temporal cortex (LTC) can be clearly appreciated.
Figure 1—figure supplement 2. Age-dependent changes of functional connectivity (FC) within the default-mode network (DMN), visual (Vis), and sensorimotor (SM) network regions during infancy.

Figure 1—figure supplement 2.

Significant age-dependent increase of within-network FC was found only in the DMN (r = 0.31, p<0.05, indicated with an asterisk) regions, but not in the Vis (r = 0.048, p=0.745) and SM (r = 0.087, p=0.559) regions.
Open in a new tab

For robust and consistent identification of functional network ROI of infants, three functional network ROIs, including DMN, visual (Vis) network, and sensorimotor (SM) network, were generated from rs-fMRI data of infant aged 12–24 months, as shown in Figure 1—figure supplement 1. After applying these network ROIs to measure functional connectivity changes of infants of all ages from 0 to 24 months, we found significant increase of the FC only within the DMN (r = 0.31, p<0.05), but not in the Vis (r = 0.048, p=0.745) or SM (r = 0.087, p=0.559), network regions (Figure 1—figure supplement 2), indicating significant functional development in the DMN, but not in the Vis or SM network.

Faster rCBF increases in the DMN hub regions during infant brain development

The labeling plane and imaging slices of pCASL perfusion MRI of a representative infant brain, reconstructed internal carotid and vertebral arteries, and four PC MR images of the target arteries are shown in Figure 2a. The rCBF maps of infant brains were calculated based on pCASL perfusion MRI and calibrated by PC MRI. As an overview, axial rCBF maps of typically developing brains at milestone ages of 1, 6, 12, 18, and 24 months are demonstrated in Figure 2b. The rCBF maps with high gray/white matter contrasts can be appreciated by a clear contrast between white matter and gray matter. A general increase of blood flow across the brain gray matter from birth to 2 years of age is readily observed. Heterogeneous rCBF distribution at a given infant age can be appreciated from these maps. For example, higher rCBF values in primary visual cortex compared to other brain regions are clear in younger infant at around 1 month. Figure 2b also demonstrates differential rCBF increases across brain regions. RCBF increases are prominent in the PCC, indicated by green arrows. On the other hand, rCBF in the visual cortex is already higher (indicated by blue arrows) than other brain regions in early infancy and increases slowly across infant development. The adopted pCASL protocol is highly reproducible with intraclass correlation coefficient (ICC) 0.8854 calculated from pCASL scans of a randomly selected infant subject aged 17.6 month, shown in Figure 2—figure supplement 1. With rCBF measured at these functional network ROIs, Figure 2—figure supplement 2 quantitatively exhibits spatial inhomogeneity of rCBF distribution regardless of age. These quantitative measurements are consistent to the observation of heterogeneous rCBF distribution in Figure 2b. Specifically, as shown in Figure 2—figure supplement 2, significant heterogeneity of rCBF was found across regions (F(6, 282) = 122.6, p<10–10) with an analysis of variance (ANOVA) test. With further paired t-test between regions, the highest and lowest rCBF was found in the Vis (82.1 ± 2.19 ml/100 g/min) and SM (49.1 ± 1.49 ml/100 g/min) regions, respectively (all ts(47) > 4.17, p<0.05, false discovery rate [FDR] corrected), while rCBF in the DMN (67.8 ± 2.08 ml/100 g/min) regions was in the middle (all ts (47) > 2.87, p<0.05, FDR corrected). Within the DMN, rCBF in the PCC (75.4 ± 2.19 ml/100 g/min) and LTC (72.0 ± 2.82 ml/100 g/min) regions were significantly higher than rCBF in the MPFC (60.7 ± 2.24 ml/100 g/min) (both ts(47) > 8.22, p<0.05, FDR corrected) and IPL regions (59.4 ± 1.96 ml/100 g/min) (both ts(47) > 7.87, p<0.05, FDR corrected). After comparing corresponding rCBF measures of different network ROIs between left and right hemisphere for evaluating rCBF asymmetry, we found significantly higher (ts(47) = 3.82, p<0.05) rCBF in the SM network ROI in the right hemisphere (50.8 ± 1.67 ml/100 g/min) compared to that in the left hemisphere (47.8 ± 1.43 ml/100 g/min) while no significant rCBF difference was found in the DMN or Vis network ROIs between two hemispheres. This finding of rCBF asymmetry in the SM network ROI is consistent to the previous studies (Chiron et al., 1997; Lemaître et al., 2021).

Figure 2. Acquisition of high-quality infant pseudo-continuous arterial-spin-labeled (pCASL) perfusion and phase contrast (PC) MRI and resultant axial regional cerebral blood flow (rCBF) maps at different infant ages.

(a) Labeling plane (red line) and imaging volume (blue box) of pCASL perfusion MRI are shown on the mid-sagittal slice of T1-weighted image of a representative infant on the left panels. Axial and sagittal view of MR angiography with reconstructed internal carotid and vertebral arteries are shown in the middle of panel (a). On the right of panel (a), the coronal view of the reconstructed arteries is placed in the middle with four slices (shown as blue bars) of the PC MR scans positioned perpendicular to the respective feeding arteries. The PC MR images are shown on the four panels surrounding the coronal view of the angiography. These PC MR images measure the global cerebral blood flow of internal carotid and vertebral arteries and are used to calibrate rCBF. (b) rCBF maps of representative typically developing (TD) infant brains at 1, 6, 12, 18, and 24 months from left to right. Axial slices of rCBF maps from inferior to superior are shown from bottom to top of the panel b for each TD infant brain. Green arrows point to the posterior cingulate cortex (a hub of the DMN network) characterized by relatively lower rCBF at early infancy and prominent rCBF increases from 1 to 24 months. Blue arrows point to the visual cortex characterized by relatively higher rCBF at early infancy and relatively mild rCBF increase from 1 to 24 months.

Figure 2.

Figure 2—figure supplement 1. Highly reproducible pseudo-continuous arterial-spin-labeled (pCASL) protocol adopted in this study for measuring regional cerebral blood flow (rCBF).

Figure 2—figure supplement 1.

Panel (a) shows reproducible rCBF maps measured from first and second half of the control/label volumes, and panel (b) shows the intraclass correlation coefficient (ICC) 0.8854 with the 95% confident intervals of [0.88, 0.8906] calculated from entire brain rCBF maps measured from first and second half control/label volumes. Both panels (a) and (b) are generated based on pCASL scans of a randomly selected infant subject aged 17.6 months.
Figure 2—figure supplement 2. Heterogeneity of regional cerebral blood flow (rCBF) measurements across functional network regions.

Figure 2—figure supplement 2.

Significant differences (p<0.05 with false discovery rate [FDR] correction) of rCBF measurements between regions were indicated by asterisks. Sample size is 48. Error bars indicate standardard deviation of rCBF measurement within each region of interests (ROI). To test significant difference of the rCBF values among different functional network ROIs, a one-way ANOVA (Analysis of Variance) with repeated measures was conducted. Paired t-tests were also conducted to test the difference of rCBF measurements between regions. FDR of each test was corrected to control the type I error. Significant interaction between regions and age was tested with an ANCOVA (Analysis of Covariance) test where age was used as a covariate.
Open in a new tab

Figure 3a shows cortical maps of linearly fitted rCBF values of infant brains from 0 to 24 months. Consistent with nonuniform profile of the rCBF maps observed in Figure 2b, the three-dimensionally reconstructed rCBF distribution maps in Figure 3a are also not uniform at each milestone infant age. RCBF increases from 0 to 24 months across cortical regions are apparent, as demonstrated by the relatively high rCBF–age correlation r values across the cortical surface in Figure 3b. Heterogeneity of rCBF increases across all brain voxels can be more clearly appreciated in Figure 3a and b compared to Figure 2b. Significant interaction between regions and age was found (F(6, 322) = 2.45, p<0.05) with an analysis of covariance (ANCOVA) test where age was used as a covariate. With DMN functional network regions including PCC, MPFC, IPL, and LTC as well as Vis and SM network regions delineated in Figure 1—figure supplement 1b as ROIs, rCBF trajectories in Figure 3c demonstrate that rCBF in these ROIs all increase significantly with age (Vis: r = 0.53, p<10–4; SM: r = 0.52, p<10–4; DMN: r = 0.7, p<10–7; DMN_PCC: r = 0.66, p<10–6; DMN_MPFC: r = 0.67, p<10–6; DMN_IPL: r = 0.66, p<10–6; DMN_LTC: r = 0.72, p<10–8). Using the trajectory of primary sensorimotor (SM) (black line and circles) in Figure 3c as a reference, rCBF increase rates across functional network ROIs are also heterogeneous (Figure 3c). Specifically, significantly higher (all p<0.05, FDR corrected) rCBF increase rate was found in total DMN ROIs (1.59 ml/100 g/min/month) and individual DMN ROIs, including DMN_PCC (1.57 ml/100 g/min/month), DMN_MPFC (1.63 ml/100 g/min/month), DMN_IPL (1.42 ml/100 g/min/month), and DMN_LTC (2.22 ml/100 g/min/month) compared to in the SM ROI (0.85 ml/100 g/min/month). Although the rCBF growth rate in the Vis (1.27 ml/100 g/min/month) ROIs is higher than that in the SM ROIs, this difference is not significant (p=0.13). Collectively, Figures 2 and 3 show that the CBF increases significantly and differentially across brain regions during infancy, with rCBF in the DMN hub regions increasing faster than rCBF in the SM and Vis regions (Figure 3). The 4D spatiotemporal whole-brain rCBF changes during infant development are presented in Video 1.

Figure 3. 4D spatiotemporal regional cerebral blood flow (rCBF) dynamics and faster rCBF increases in the default-mode network (DMN) hub regions during infancy.

Figure 3.

Open in a new tab

(a) Medial (top row) and lateral (bottom row) views of fitted rCBF profiles of the infant brain at 0, 6, 12, 18, and 24 months in the custom-made infant template space demonstrate heterogeneous rCBF increase across the brain regions. (b) Medial (top) and lateral (bottom) views of rCBF–age correlation coefficient (r) map are demonstrated. (c) The scatterplots of rCBF measurements in the primary sensorimotor (SM) network (black circle and black line), visual (Vis) network (blue circle and blue line), and total and individual DMN hub regions (DMN_MPFC, DMN_PCC, DMN_IPL, and DMN_LTC) (blue circle and blue line) of all studied infants demonstrate differential rCBF increase rates. * next to network name in each plot indicates significant (false discovery rate [FDR]-corrected p<0.05) differences of rCBF trajectory slopes from that of SM used as a reference and shown in a black dashed line. See legend of Figure 1 for abbreviations of the DMN subregions.

Video 1. Video of the 4D spatiotemporal whole-brain dynamics of regional cerebral blood flow from 0 to 24 months.

Download video file (825.6KB, mp4)
Open in a new tab

Coupling between rCBF and FC within DMN during infant brain development

To test the hypothesis that rCBF increases in the DMN regions underlie emergence of this vital functional network, correlation between rCBF and FC was conducted across randomly selected voxels within the DMN of all infants aged 0–12 months (Figure 4a) and all infants aged 12–24 months (Figure 4b). Significant correlations (p<0.001) were found in both age groups. Partial correlation analysis between rCBF and FC after regressing out age effects also confirmed significant correlation between rCBF and FC in the DMN regions in both 0–12-month (p<0.001) and 12–24-month (p<0.001) groups excluding the age effect. We further tested whether functional emergence of the DMN represented by increases of FC within the DMN (namely DMN FC) was correlated to rCBF increases specifically in the DMN regions, but not in primary sensorimotor (Vis or SM) regions. Figure 5a shows correlations between the DMN FC and rCBF at the DMN (red lines), Vis (green lines), or SM (blue lines) voxels. The correlations between the DMN FC and averaged rCBF in the DMN, Vis, or SM region are represented by thickened lines in Figure 5a. A correlation map (Figure 5b) between the DMN FC and rCBF across the entire brain voxels was generated. The procedures of generating this correlation map are illustrated in Figure 5—figure supplement 1. The DMN, Vis, and SM ROIs in Figure 5b were delineated with dashed red, green, and blue contours, respectively, and obtained from Figure 1—figure supplement 1b. Most of the significant correlations (r > rcrit) between the DMN FC and voxel-wise rCBF were found in the voxels in the DMN regions, such as PCC, IPL, and LTC, but not in the Vis or SM regions (Figure 5b). Demonstrated in a radar plot in Figure 5c, much higher percent of voxel with significant correlations between rCBF and the DMN FC was found in the DMN (36.7%, p<0.0001) regions than in the SM (14.6%, p>0.05) or Vis (5.5%, p>0.05) regions. Statistical significance of higher percent of voxels with significant correlations in the DMN (p<0.0001) was confirmed using nonparametric permutation tests with 10,000 permutations. We also conducted the correlation between the Vis FC and rCBF across the brain as well as permutation test. As expected, no significant correlation between the Vis FC and rCBF can be found in any voxel in the DMN, Vis, or SM ROIs, demonstrated in Figure 5—figure supplement 2a. Similar analysis was also conducted for correlation between the SM FC and rCBF across the brain and percent of voxels with significant correlation was close to zero, as demonstrated in Figure 5—figure supplement 2b. Combined with the results shown in Figure 5, the results of coupling between Vis (Figure 5—figure supplement 2a) or SM (Figure 5—figure supplement 2b) FC and rCBF further demonstrated that the selected rCBF-FC coupling can be only found in the DMN ROIs, but not in the Vis or SM network ROIs.

Figure 4. Significant correlation of regional cerebral blood flow (rCBF) and functional connectivity (FC) at randomly selected 4000 voxels within the default-mode network (DMN) for both infants aged 0–12 months (p<0.001, left scatter plot) and infants aged 12–24 months (p<0.001, right scatter plot).

Figure 4.

Open in a new tab

FC is the average of FC of a certain DMN voxel to all other DMN voxels. The DMN regions of interests obtained from a data-driven independent component analysis of resting-state fMRI of the 12–24month infant cohort are shown on the top panels as an anatomical reference. See legend of Figure 1 for abbreviations of the DMN subregions.

Figure 5. Significant correlation between functional emergence of the default-mode network (DMN) and regional cerebral blood flow (rCBF) increases specifically in the DMN regions, but not in primary sensorimotor (visual or sensorimotor) regions.

(a) Correlation of intra-default-mode-network functional connectivity (DMN FC) and rCBF at randomly selected voxels in the DMN (light red lines), visual (Vis, light green lines) and sensorimotor (SM, light blue lines) network regions. Correlations of DMN FC and averaged rCBF in the DMN, Vis, and SM network regions are shown as thickened red, green, and blue lines, respectively. (b) Coupling between the DMN FC and rCBF across the brain can be appreciated by distribution of voxel-wise correlation coefficient (r) obtained from correlation between DMN FC and rCBF at each voxel. The short black line in the color bar indicates critical r value rcrit corresponding to p=0.05. Higher r values can be appreciated in the DMN hub regions including posterior cingulate cortex (PCC), medial prefrontal cortex (MPFC), inferior posterior lobule (IPL), and lateral temporal cortex (LTC) with their boundaries delineated by the dashed dark red contours (from Figure 1—figure supplement 1b). Dashed green and blue contours (also from Figure 1—figure supplement 1b) delineate the Vis and SM network regions, respectively. (c) Radar plot shows significant correlation between rCBF and intra-DMN FC in the DMN network (36.7%, p<0.0001), but not in the Vis (14.6%, p>0.05), or SM (5.5%, p>0.05) networks. The radius represents the percent of the voxels with significant correlations between intra-DMN FC and rCBF in DMN, Vis, and SM network regions, respectively. The dashed line circle indicates critical percent of significant voxels with p=0.05 from 10,000 permutation tests. ***p<0.0001.

Figure 5.

Figure 5—figure supplement 1. Procedures of generating map of correlation between the default-mode network (DMN) functional connectivity (FC) and reginal cerebral blood flow (rCBF) across the entire brain voxels.

Figure 5—figure supplement 1.

At any brain voxel indicated by a small black box in the correlation map below, correlation between the DMN FC across subjects and the rCBF at this voxel across subjects was conducted and the correlation coefficient was calculated for this voxel. The correlation map between the DMN FC and rCBF across the brain can be generated by projecting the correlation coefficient r of each voxel onto the cortical surface.
Figure 5—figure supplement 2. Coupling between intra-visual network functional connectivity (Vis FC) and regional cerebral blood flow (rCBF) and coupling between intra-sensorimotor network (SM FC) and rCBF.

Figure 5—figure supplement 2.

Left panels show the correlation of Vis FC (a) or SM FC (b) and rCBF at randomly selected voxels in the default-mode (DMN) (light red lines), Vis (light green lines) and SM (light blue lines) network regions. Correlations of Vis FC (a) or SM FC (b) and averaged rCBF in the DMN, Vis, and SM network regions are shown as thickened red, green, and blue lines, respectively. Middle panels demonstrate the distribution of voxel-wise correlation coefficient (r) obtained from correlation between Vis FC (a) or SM FC (b) and rCBF at each voxel. The short black line in the color bar indicates critical correlation coefficient r value rcrit corresponding to p=0.05. Low r values are apparent in all network (DMN, Vis, and SM) regions with dashed red, green, and blue contours (from Figure 1—figure supplement 1b) delineating DMN, Vis, and SM network, respectively. Right panels show radar plot with radius representing percent of voxels with significant correlation between Vis FC or SM FC and rCBF in the DMN, Vis, or SM network regions. The radii in radar plots demonstrate that the percent of voxels with significant correlations between intra-Vis (a) or intra-SM (b) FC and rCBF in DMN, Vis, or SM network regions is 0 or close to 0, indicating that coupling between Vis FC or SM FC and rCBF in any network region is not significant. The dashed line circle in the radar plot indicates critical percent of significant voxels with p=0.05 from 10,000 permutation tests.
Figure 5—figure supplement 3. A diagram illustrating hypothesized neuronal mechanism supporting coupling of regional cerebral blood flow (rCBF) and functional connectivity.

Figure 5—figure supplement 3.

Specifically, higher cerebral blood flow delivers larger amount of oxygen and glucose to neurons, leading to cascade of events of regional cerebral metabolic rate of oxygen (CMRO2) ↑ and glucose (CMRGlu) ↑, synaptogenesis ↑ and synaptic efficacy ↑, blood oxygenation level-dependent (BOLD) signal synchronization ↑, and functional connectivity ↑, during infant brain development.
Open in a new tab

Discussion

We revealed strongly coupled rCBF and FC increases specifically in the DMN while establishing unprecedented 4D rCBF spatiotemporal changes during infancy. The tight rCBF-FC relationship found with multimodal infant MRI suggests that DMN emergence is supported by faster local blood flow increase in the DMN to meet metabolic demand, offering refreshing insight into the physiological mechanism underlying early brain functional architecture emergence. The delineated 4D brain perfusion spatiotemporal framework was characterized with heterogeneous rCBF distribution across brain regions at a specific age and differential age-dependent rCBF increase rates across brain regions during infant development, and can be used as quantified standard reference for detecting rCBF alterations (e.g., the z scores) of atypically developing brains. Elucidating the ontogeny of infant brain physiology and its functional correlates could greatly advance current understanding of general principles of early brain development.

Gradient of functional network maturations in early brain development has been more extensively characterized with recent rs-fMRI studies. Differential emergence of these functional networks is distinguished by different onset time as well as different maturational rate of various brain functions in a given developmental period. For example, primary sensory and motor functional networks, such as the SM and Vis networks, appear earlier before or around birth (Cao et al., 2017a; Doria et al., 2010; Fransson et al., 2007; Smyser et al., 2010; Peng et al., 2020). Other functional networks involved in heteromodal functions appear later. The DMN (Fox et al., 2007; Greicius et al., 2003; Greicius et al., 2009; Raichle et al., 2001; Raichle, 2015; Smith et al., 2009) is a higher-order functional network. Smyser et al., 2010 found that SM and Vis functional networks mature earlier and demonstrate adult-like pattern for preterm neonate brain, with the DMN much immature and incomplete around birth.Cao et al., 2017a also found rapid maturation of primary sensorimotor functional systems in preterm neonates from 31 to 41 postmenstrual weeks while the DMN remained immature during that period. These previous studies suggest that significant functional maturation in primary sensorimotor networks occur earlier in preterm and perinatal developmental period (Cao et al., 2017a; Doria et al., 2010; Smyser et al., 2010) compared to 0–24-month infancy focused in this study. Functional network emergence in the DMN was found in the developmental infant cohort in Figure 1, marking significant maturation of the DMN in infancy and distinguished network pattern from earlier developmental period. The delineated DMN emergence in this study is also consistent to the literature (Gao et al., 2009). Figure 1—figure supplement 2 further demonstrated significant increase of FC only in the DMN, but not in primary sensorimotor system that already emerged in earlier developmental period.

Glucose and oxygen are two primary molecules for energy metabolism in the brain (Raichle et al., 2001; Vaishnavi et al., 2010). Glucose consumed by infant brain represents 30% total amount of glucose (Raichle, 2010; Settergren et al., 1976), more than 15–20% typically seen in adult brain (Bouma and Muizelaar, 1990; Satterthwaite et al., 2014). The cerebral metabolic rates for glucose (CMRGlu) and oxygen (CMRO2) are direct measures of the rate of energy consumption, which parallel the proliferation of synapses in brain during infancy (Raichle, 2010). RCBF delivering glucose and oxygen for energy metabolism in the brain is closely related to CMRGlu and CMRO2 and can serve as a surrogate of these two measurements (Fox and Raichle, 1986; Gur et al., 2009; Paulson et al., 2010; Vaishnavi et al., 2010). In the PET study (Chugani and Phelps, 1986) using CMRGlu measurements, it was found that the local CMRGlu in the sensorimotor cortex almost reaches the highest level in early infancy and then plateaus during rest of infancy, consistent with relatively small changes of rCBF in later infancy in primary sensorimotor ROIs found in this study (Figure 3). On the other hand, the global CBF measured with PC MRI (Liu et al., 2019) increases dramatically during infancy with global CBF at 18 months almost five times of the CBF around birth. Taken together, the literature suggests significant but nonuniformly distributed CBF increases across the brain regions during infancy, consistent to the measured heterogeneous rCBF increase pattern (Figures 2 and 3) in this study.

Furthermore, the differentiated cerebral metabolic pattern reflected by measured rCBF distribution (Figure 2) at a specific age and differential increase rates of the rCBF (Figure 3) from 0 to 24 months are strikingly consistent with spatiotemporally differentiated functional (Cao et al., 2017a; Doria et al., 2010; Fransson et al., 2007; Gao et al., 2009; Smyser et al., 2010) and structural (Ouyang et al., 2019a) maturational processes. In developmental brains, cellular processes supporting differential functional emergence require extra oxygen and glucose delivery through cerebral blood flow to meet the metabolic demand. In Hebb’s principle, 'neurons that fire together wire together.' Through the synaptogenesis in neuronal maturation, the neurons within a certain functional network system tend to have more synchronized activity in a more mature stage than in an immature stage. As shown in the diagram in Figure 5—figure supplement 3, cellular activities in developmental brains, such as synaptogenesis critical for brain circuit formation, need extra energy more than that in the stable and matured stage. Neurons do not have internal reserves of energy in the form of sugar or oxygen. The demand of extra energy requires an increase in rCBF to deliver more oxygen and glucose for the formation of brain networks. In the context of infant brain development, there is a cascade of events of CBF increase, CMRO2 and CMRGlu increase, synaptogenesis and synaptic efficacy increase, blood oxygenation level-dependent (BOLD) signal synchronization increase, and functional connectivity increase, shown in the bottom of Figure 5—figure supplement 3. Such spatial correlation of rCBF to the FC in the functional network ROIs was found in adults (Liang et al., 2013). Higher rCBF has also been found in the DMN in children 6–20 years of age (Liu et al., 2018). Consistent with the diagram shown in Figure 5—figure supplement 3, Figure 4 revealed significant correlation between FC and rCBF in the DMN network, and Figure 5 identified this significant correlation between FC and rCBF specifically in the DMN network, but not in the primary Vis or SM networks. Collectively, these results (Figures 4 and 5) were well aligned with the hypothesis that faster rCBF increase in the DMN underlies the emergence of the DMN reflected by significant FC increases. The revealed physiology–function relationship may shed light on physiological underpinnings of brain functional network emergence.

It is noteworthy to highlight a few technical details below. First, this study benefits from multimodal MRI allowing measuring functional network emergence and rCBF of the same cohort of infants. Simultaneous rCBF and FC measurements enabled us to probe relationship of brain physiology and function during infant development. Second, a nonparametric permutation analysis without a prior hypothesis at the voxel level across the whole brain was conducted to confirm the coupling of rCBF and FC is specific in the DMN regions (Figure 5), not in the primary sensorimotor (Vis or SM) regions (Figure 5—figure supplement 2), demonstrating the robustness of the results on physiological underpinning of functional network emergence in the DMN. Third, the pCASL perfusion MRI was calibrated by PC MRI so that the errors caused by varying labeling efficiency and varying T1 value of arterial blood among infant subjects can be ameliorated. This subject-specific calibration process therefore enhanced the accuracy of rCBF measurements for their potential use as a standard reference. For studies without additional PC MRI acquired for subject-specific calibration, utilizing developmentally appropriate or even subject-specific T1 values of arterial blood is encouraged to improve the accuracy of rCBF measurements. Given the lack of consensus on pCASL acquisition parameters for infants, future research focusing on systematically optimizing the pCASL acquisition protocol in the infant population is needed. Fourth, the ROIs for rCBF measurements were obtained by data-driven ICA of the same sub-cohort of infant subjects aged 12–24 months instead of transferred ROIs from certain brain parcellation atlases. Since most of the parcellation atlases were built based on adult brain data and all these atlases were established based on other subject groups, ROIs delineated from the same cohort improve accuracy of coupling analysis. There are several limitations in this study that can be improved in future investigations. All data were acquired from a cross-sectional cohort. To minimize the inter-subject variability, future study with a longitudinal cohort of infants is warranted. With relatively small size of infant brains, spatial resolution of the pCASL and rs-fMRI can be further improved too to improve imaging measurement accuracy. With relatively larger voxel size of the pCASL scan, partial volume effects could lead to bias of the rCBF measurement. Due to smaller brain and thinner brain cortex of younger infants, using consistent pCASL acquisition voxel size could lead to heterogeneous partial volume effects across infants with different brain sizes and cortical thicknesses. The rCBF biases related to partial volume effects were minimized with calibration of individual PC MRI from all infants. Relatively larger pCASL acquisition voxel size might also contribute to blurring the smaller red areas in the occipital lobe of the brain into a larger continuous red region, although higher rCBF values at the base of the brain and occipital regions than in the frontal regions were observed in infant and neonate brains (e.g., Kim et al., 2018; Lemaître et al., 2021; Wang et al., 2008; Satterthwaite et al., 2014). We believe that the rCBF increase pattern during maturation is more complicated than relatively simplified posterior-to-anterior and inferior-to-superior gradients. It also presents a primary-to-association gradient reproducibly found in the literature (see Sydnor et al., 2021 for review). Such mixed patterns are consistent with the maturation pattern demonstrated in Figure 3a. More clear maturation gradient could benefit from future rCBF maps of higher signal-to-noise ratio and higher resolution. To further improve the statistical power, larger infant sample size will be beneficial in the future studies. Finally, although the rCBF from pCASL perfusion MRI is highly correlated with the CMRO2 and CMRGlu measured from PET, rCBF is not a direct measurement of the rate of energy consumption. Physiology–function relationship studies in infants could benefit from the development of novel noninvasive MR imaging methods as an alternative to PET to measure CMRO2 or CMRGlu without the involvement of radioactive tracer.

Conclusion

Novel findings in this study inform a physiological mechanism of DMN emergence during infancy with rCBF and FC measurements from multimodal MRI in developmental infant brains. The age-specific whole-brain rCBF maps and spatiotemporal rCBF maturational charts in all brain regions serve as a standardized reference of infant brain physiology for precision medicine. The rCBF-FC coupling results revealing fundamental physiology–function relationship have important implications in altered network maturation in developmental brain disorders.

Materials and methods

Infant subjects

Forty-eight infants (30 males) aged 0–24 months (14.6 ± 6.32 months) were recruited at Beijing Children’s Hospital. These infants were referred to MRI due to simple febrile convulsions (n = 38), diarrhea (n = 9), or sexual precocity (n = 1). All infants had normal neurological examinations documented in medical record. The exclusion criteria include known nervous system disease or history of neurodevelopmental or systemic illness. Every infant’s parents provided signed consent, and the protocol was approved by the Beijing Children’s Hospital Research Ethics Committee (approval number 2016-36).

Data acquisition

All infant MR scans, including rs-fMRI, pCASL, PC MRI, and structural MRI, were acquired with the same 3T Philips Achieva system under sedation by orally taken chloral hydrate with dose of 0.5 ml/kg and no more than 10 ml in total. Previous studies (Li et al., 2011; Suzuki et al., 2021) suggested no significant impact of chloral hydrae on CBF or sensory function. Earplug and headphones were used to minimize noise exposure. Resting state fMRI (rs-fMRI) images were acquired using echo planar imaging with the following parameters: TR = 2000 ms, TE = 24 ms close to TE used for adults since the relative BOLD parameter of infants aged 9 months was found quite close to that of adults (Cusack et al., 2018), flip angle = 60°, 37 slices, FOV = 220 × 220 mm2, matrix size = 64 × 64, voxel size = 3.44 × 3.44 × 4 mm3. 200 dynamics were acquired for each infant. The acquisition time of the rs-fMRI images was 7 min. Pseudo-continuous arterial spin labeling (pCASL) perfusion MRI images were acquired using a multi-slice echo planar imaging with the following parameters: TR = 4100 ms, TE = 15 ms, 20 slices with 5 mm slice thickness and no gap between slices, field of view (FOV) = 230 × 230 mm2, matrix size = 84 × 84, voxel size = 2.74 × 2.74 × 5 mm3. As shown on the left panel of Figure 2a, the labeling slab was placed at the junction of spinal cord and medulla (65 mm below central slab of imaging volume) and parallel to the anterior commissure-posterior commissure (AC-PC) line. The labeling duration was 1650 ms, and the post labeling delay (PLD) was 1600 ms. Thirty pairs of control and label volume were acquired for each infant. The acquisition time of the pCASL perfusion MRI images was 4.2 min. An auxiliary scan with identical readout module to pCASL but without labeling was acquired for estimating the value of equilibrium magnetization of brain tissue. For accurate and reliable voxel-wise comparison of rCBF across different infant ages, consistent imaging parameters were applied in this study. Imaging parameters, including PLD of infants, were selected closer to those of children and in the interval between imaging parameters of neonates and those of children, with PLD 2000 ms for neonates and 1500 ms for children recommended by the white paper (Alsop et al., 2015). The mean arterial transit time (ATT) in infants was reported to be less than 1500 ms (Varela et al., 2015; Kim et al., 2018). Also following the ASL white paper (Alsop et al., 2015) suggestion that the selected PLD should be larger than the ATT, a tailored and consistent PLD of 1600 ms larger than ATT of infants was selected for the studied infant cohort to minimize the effect of PLD on rCBF differences across subjects or across brain regions. PC MRI was acquired to calibrate rCBF by scaling the overall CBF of entire brain. The carotid and vertebral arteries were localized based on a scan of time-of-flight (TOF) angiography acquired with the following parameters: TR = 20 ms, TE = 3.45 ms, flip angle = 30°, 30 slices, FOV = 100 × 100 mm2, matrix size = 100 × 100, voxel size = 1 × 1 × 1 mm3. The acquisition time of the TOF angiography was 28 s. Based on the angiography, the slices for the PC MRI of internal carotid arteries were placed at the level of the foramen magnum and the slices for the PC MRI of vertebral arteries were placed between the two turns in V3 segments, as illustrated on the right panel of Figure 2a. For left and right internal carotid and vertebral arteries, PC MRI images were acquired with the following parameters: TR = 20 ms, TE = 10.6 ms, flip angle = 15°, single slice, FOV = 120 × 120 mm2, matrix size = 200 × 200, voxel size = 0.6 × 0.6 × 3 mm, maximum velocity encoding = 40 cm/s, non-gated, four repetitions. The acquisition time of PC MRI for each artery was 24 s. T1-weighted images of all infants were also acquired for anatomical information and brain segmentation using MPRAGE (Magnetization Prepared – RApid Gradient Echo) sequence with the following parameters: TR = 8.28 ms, TE = 3.82 ms, flip angle = 12°, TI (time of inversion) = 1100 ms, 150 slices, FOV = 200 × 200 mm2, matrix size = 200 × 200, voxel size = 1 × 1 × 1 mm3. The acquisition time of the T1-weighted image was 3.7 min. Visual inspection was carefully conducted for all MRI data by experienced pediatric radiologists (DH and YP) with decades of experience in clinical radiology. No significant motion artifacts were spotted with the sedated MRI scans. Therefore, no dataset from the 48 infants was excluded from the following data analysis due to severe motion artifacts.

Rs-fMRI preprocessing

The same preprocessing procedures elaborated in our previous publication (Cao et al., 2017b) was used. Briefly, the normalized rs-fMRI images underwent spatial smoothing with a Gaussian kernel of full width at half-maximum (FWHM) of 4 mm, linear trend removal, and temporal band-pass filtering (0.01–0.10 Hz). Several nuisance variables, including six rigid-body head motion parameters and the averaged signal from white matter and cerebrospinal fluid (CSF) tissue, were removed through multiple linear regression analysis to reduce the effects of non-neuronal signals. The CSF and white matter were segmented with the T1-weighted image using SPM. Preprocessed rs-fMRI signals were used to estimate functional connectivity (FC), defined as Pearson’s correlation between the time courses of preprocessed (rs-fMRI BOLD) signal in two regions or two voxels (Figure 1—figure supplement 1a).

Identification of functional network ROIs with rs-fMRI

Brain regions of three functional networks, including DMN, Vis network, and SM network, were used as functional ROIs for quantifying rCBF. These functional networks were identified with rs-fMRI of 35 infants aged 12–24 months as DMN can be better delineated in later infancy than earlier infancy. ICA (Beckmann et al., 2005) in the FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/MELODIC) was used to identify all these network regions. Individual DMN regions, including PCC, MPFC, IPL, and LTC, were extracted from the independent components (ICs) based on the spatially distributed regions consistently identified across studies (Greicius et al., 2004; Smith et al., 2009). Vis and SM network regions were also identified following the literature (Smith et al., 2009). The functional network ROIs were obtained by thresholding the ICs at z value of 1.96, which corresponds to p value of 0.05. The identified ROIs for DMN, Vis, and SM networks are shown in Figure 1—figure supplement 1b. ROIs of DMN, Vis, or SM include all regions in each network, respectively. The functional network ROIs overlapped with the binary cortical mask in the template space were used for measuring ROI-based rCBF below.

Measurement of rCBF with pCASL perfusion MRI and calibrated by PC MRI

After head motion correction of the pCASL perfusion MRI, we estimated rCBF using the protocol similar to that in our previous publication (Ouyang et al., 2017). Briefly, rCBF was measured using a model described in ASL white paper (Alsop et al., 2015):

rCBF=6000⋅λ⋅ΔM⋅ePLDT1a2⋅α⋅Mb0⋅T1a⋅(1−e−LabelDurT1a)[ml/100g/min] (1)

where ∆M is the dynamic-averaged signal intensity difference between in the control and label images; λ, the blood–brain partition coefficient, is 0.9 ml/g (Herscovitch and Raichle, 1985); PLD, the post labeling delay time, is the cumulation of 1650 ms and the delayed time between slices; LabelDur, the labeling duration, is 1600 ms; α, the labeling efficiency, is 0.86 predicted by the fitting between labeling efficiency and blood velocity in the previous study (Aslan et al., 2010); T1a, T1 of arterial blood, is 1800 ms (Liu et al., 2016; Varela et al., 2011). The value of equilibrium magnetization of brain tissue (Mb0) was obtained from an auxiliary scan with identical readout module of pCASL except labeling. The labeling efficiency α can vary considerably across participants, especially in infants. Thus, we used PC MRI to estimate and calibrate rCBF measures, as described previously (Aslan et al., 2010; Ouyang et al., 2017). To calibrate rCBF, global CBF from PC MRI was calculated as follows:

fPC,AVG=∫vdA/(ρ∗brainvolume) (2)

where v is the blood flow velocity in the ICAs and VAs; A is the cross-sectional area of the blood vessel with the unit mm2; and the brain tissue density ρ is assumed as 1.06 g/mL (Dittmer and Dawson, 1961; Herscovitch and Raichle, 1985). Brain volume was measured from the T1-weighted image as parenchyma volume (gray matter + white matter volume). RCBF was calibrated by applying the scalar factor making averaged rCBF equal to global CBF from PC MRI. To demonstrate the reproducibility of the adopted pCASL protocol, the ICC was calculated based on entire brain rCBF maps measured from first half and second half of control/label series of pCASL scan of a randomly selected infant subject aged 17.6 months.

Multimodal image registration to a customized template space from all subjects

For integrating perfusion MRI and rs-fMRI data of all infant subjects, a customized structural template was generated. T1-weighted image of a 12-month-old brain characterized by median brain size at this age and straight medial longitudinal fissure was used as a single-subject template. T1-weighted images of all subjects were registered to the single-subject template by using nonlinear registration in Statistical Parametric Mapping (SPM 8, http://www.fil.ion.ucl.ac.uk/spm). The averaged T1-weighted image in the template space was defined as the structural template. In individual space, segmentation of brain gray matter was also conducted using the contrasts of T1-weighted image with SPM, generating the gray matter tissue probability map in individual space. After slice timing and head motion correction, intra-subject registration of rs-fMRI to T1-weighted image in the individual space was conducted by transforming the averaged images across dynamics in each session to the T1-weighted image of the same subject through linear registration with SPM. This intra-subject transformation was applied to each volume of rs-fMRI. As described in the main text, rCBF map was estimated in the individual space using Eq (1). Intra-subject registration of rCBF map to T1-weighted image was also conducted through linear registration with SPM. RCBF maps, gray matter probability map, and rs-fMRI images aligned to T1-weighted images in the individual space of each subject were then normalized to the customized infant template space through the same nonlinear registration from individual T1-weighted image to the structural template. After nonlinear inter-subject normalization, rCBF map was smoothed spatially with Gaussian kernel of FWHM 4 mm in the customized template space. An averaged gray matter probability map was also generated after inter-subject normalization. After carefully testing multiple thresholds in the averaged gray matter probability maps, 40% probability minimizing the contamination of white matter and CSF while keeping the continuity of the cortical gray matter mask was used to generate the binary cortical mask.

Characterization of age-dependent changes of rCBF and FC

After multimodal images from all subjects were registered in the same template space, age-dependent rCBF and FC changes were characterized using linear regression. The linear cross-sectional developmental trajectory of voxel-wise rCBF increase was obtained by fitting rCBF measurements in each voxel with ages across subjects. The fitted rCBF at different ages was estimated and projected onto the template cortical surface with Amira (FEI, Hillsboro, OR) to show the spatiotemporal changes of rCBF across cortical regions during infancy. The rCBF–age correlation coefficients r values of all voxels were estimated and mapped to the cortical surface, resulting in the correlation coefficient map. The functional ROIs identified above were used to quantify regional rCBF change. The rCBF in the SM, Vis, and DMN ROIs were averaged across voxels within each ROI, and fitted with a linear model: rCBF(t) = α + β t + ε, where α and β are intercepts and slopes for rCBF measured at certain ROI, t is the infant age in months, and ε is the error term. To test whether regional rCBF increased significantly with age, the null hypothesis was that rCBF slope in each ROI was equal to zero. To compare regional rCBF change rates between the tested ROI and SM ROI, the null hypothesis was that the rCBF slope of the tested ROI and rCBF slope of SM ROI were equal. Rejection of the null hypothesis indicated a significant rCBF slope difference between two ROIs.

Nonlinear models, including exponential and biphasic models, were compared with linear model using F-test for fitting regional rCBF in the DMN, SM, and Vis ROIs. No significant difference was found between linear and exponential (all F(1,45) < 3.45, p>0.05, uncorrected) fitting or between linear and biphasic (all F(2,44) < 2.64, p>0.05, uncorrected) fitting in the regional rCBF changes in the DMN, SM, and Vis ROIs.

In the customized template space, time-dependent changes of FC to PCC and time-dependent changes of FC within the functional network ROIs were calculated. FC of any given voxel outside the PCC to PCC was defined as correlation between signal time course of this given voxel and the averaged signal time course of all voxels within the PCC. Emergence of the DMN was delineated by linear fitting of the age-related increase of FC to PCC in each voxel across subjects. FC within each network was defined as the mean of functional connectivity strength (Cao et al., 2017b) of each voxel within this network region. The FC within the DMN, SM, or Vis was also fitted with a linear model: FC(t) = α + β t + ε, where α and β are intercepts and slopes for FC measured within a certain network, t is the infant age in months, and ε is the error term. To test whether the FC within a network increased significantly with age, the null hypothesis was that rCBF slope in each ROI was equal to zero.

Test of heterogeneity of rCBF across functional network ROIs

To examine heterogeneity of infant rCBF in different brain regions, rCBF measurements of all infants were averaged across voxels in the Vis, SM, and DMN (including DMN subregions DMN_PCC, DMN_MPFC, DMN_IPL, and DMN_LTC), respectively. To test significant difference of the rCBF values among different functional network ROIs, a one-way ANOVA with repeated measures was conducted. Paired t-tests were also conducted to test the difference of rCBF measurements between regions. FDR of each test was corrected to control the type I error. Significant interaction between regions and age was tested with an ANCOVA test where age was used as a covariate.

Coupling between rCBF and FC during the infant brain development

Coupling between rCBF and FC in the DMN was conducted with voxel-wise approach. The FC of a voxel in the DMN was the average of correlations of rs-fMRI BOLD signal between this voxel and all other DMN voxels. All infants were divided into two groups based on their ages, 0–12 months and 12–24 months. 4000 voxels were randomly chosen from the DMN voxels of all subjects in each age group for correlation analysis. Since the variance of both FC and rCBF cannot be ignored in this study, Deming regression (Deming, 1943) was used to fit the trendline of coupling between FC and rCBF. Partial correlation between rCBF and FC regressing out age effects was also conducted to confirm the significant rCBF-FC coupling in the DMN excluding the age effects. We further tested whether significant FC-rCBF coupling was specifically localized in the DMN, but not in primary sensorimotor (Vis or SM) regions. FC within a specific network was calculated by averaged FC of all voxels in this network. As demonstrated in Figure 5—figure supplement 1, the correlation between FC within a network and the rCBF at each voxel resulted in a whole-brain r map (e.g., Figure 5b). A nonparametric permutation test was then applied to evaluate the significance of rCBF–FC correlation in the ROIs of a specific brain network (e.g., DMN, SM, or Vis). The null hypothesis is that the voxels with significant correlation (r > 0.28) between rCBF and FC are distributed evenly in the brain. To test the null hypothesis, we resampled the correlation coefficient r of all brain voxels randomly for 10,000 times to build 10,000 whole-brain correlation coefficient distribution maps. The FC–rCBF correlation was considered significant in certain brain network ROIs if the number of observed significant voxels in the network ROIs is higher than the number of significant voxels corresponding to 95th percentile in the permutation tests.

Acknowledgements

This work was supported by grants from the National Institute of Health: R01MH092535, R01MH125333, R01EB031284, R21MH123930, and P50HD105354.

Funding Statement

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

Contributor Information

Hao Huang, Email: huangh6@email.chop.edu.

Timothy E Behrens, University of Oxford, United Kingdom.

Timothy E Behrens, University of Oxford, United Kingdom.

Funding Information

This paper was supported by the following grants:

  • National Institutes of Health R01EB031284 to Hao Huang.

  • National Institutes of Health R21MH123930 to Minhui Ouyang.

  • National Institutes of Health R01MH125333 to Hao Huang.

  • National Institutes of Health R01MH092535 to Hao Huang.

  • National Institutes of Health P50HD105354 to Hao Huang.

  • National Institutes of Health R01MH129981 to Hao Huang.

Additional information

Competing interests

No competing interests declared.

Author contributions

Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing.

Software, Formal analysis, Validation, Visualization, Methodology, Writing - original draft, Writing - review and editing.

Methodology, Writing - review and editing.

Resources, Data curation, Project administration, Writing - review and editing.

Resources, Data curation, Project administration, Writing - review and editing.

Methodology, Writing - review and editing.

Methodology, Writing - review and editing.

Resources, Data curation, Project administration, Writing - review and editing.

Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing.

Ethics

Informed parental consents were obtained from the subject's parent. The Institutional Review Board of Beijing Children's Hospital Research Ethics Committee (Approval number 2016-36) approved the study procedures.

Additional files

MDAR checklist

Data availability

Maps including regional cerebral blood flow (rCBF) maps and functional connectivity maps from multimodal infant MRI datasets including pseudo-continuous arterial-spin-labelled perfusion MRI and resting-state MRI of forty-eight infants are publicly available from Huang lab GitHub repository (https://github.com/haohuanglab/infant_perfusion_function, copy archived at swh:1:rev:34071e49232d7960ff6b78464b55077da136d868). However, we cannot openly share the raw, unprocessed MRI data, because The Institutional Review Board of Beijing Children's Hospital Research Ethics Committee (Approval number 2016-36) specifies the participants did not give consent for these data to be released publicly. The raw MRI data can be made available to individual researchers on informal request to the corresponding author through email hao.huang@pennmedicine.upenn.edu. Source code used in analysis and related documentation are also available at the Huang lab GitHub repository (https://github.com/haohuanglab/infant_perfusion_function).

References

  1. Alsop DC, Detre JA, Golay X, Günther M, Hendrikse J, Hernandez-Garcia L, Lu H, MacIntosh BJ, Parkes LM, Smits M, van Osch MJP, Wang DJJ, Wong EC, Zaharchuk G. Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: a consensus of the ISMRM perfusion Study Group and the European Consortium for ASL in dementia. Magnetic Resonance in Medicine. 2015;73:102–116. doi: 10.1002/mrm.25197. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Altman DI, Powers WJ, Perlman JM, Herscovitch P, Volpe SL, Volpe JJ. Cerebral blood flow requirement for brain viability in newborn infants is lower than in adults. Annals of Neurology. 1988;24:218–226. doi: 10.1002/ana.410240208. [DOI] [PubMed] [Google Scholar]
  3. Altman DI, Perlman JM, Volpe JJ, Powers WJ. Cerebral oxygen metabolism in newborns. Pediatrics. 1993;92:99–104. doi: 10.1542/peds.92.1.99. [DOI] [PubMed] [Google Scholar]
  4. Aslan S, Xu F, Wang PL, Uh J, Yezhuvath US, van Osch M, Lu H. Estimation of labeling efficiency in pseudocontinuous arterial spin labeling. Magnetic Resonance in Medicine. 2010;63:765–771. doi: 10.1002/mrm.22245. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Beckmann CF, DeLuca M, Devlin JT, Smith SM. Investigations into resting-state connectivity using independent component analysis. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences. 2005;360:1001–1013. doi: 10.1098/rstb.2005.1634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bouma GJ, Muizelaar JP. Relationship between cardiac output and cerebral blood flow in patients with intact and with impaired autoregulation. Journal of Neurosurgery. 1990;73:368–374. doi: 10.3171/jns.1990.73.3.0368. [DOI] [PubMed] [Google Scholar]
  7. Buckner RL, Andrews-Hanna JR, Schacter DL. The brain’s default network: anatomy, function, and relevance to disease. Annals of the New York Academy of Sciences. 2008;1124:1–38. doi: 10.1196/annals.1440.011. [DOI] [PubMed] [Google Scholar]
  8. Bystron I, Blakemore C, Rakic P. Development of the human cerebral cortex: Boulder Committee revisited. Nature Reviews. Neuroscience. 2008;9:110–122. doi: 10.1038/nrn2252. [DOI] [PubMed] [Google Scholar]
  9. Cao M, He Y, Dai Z, Liao X, Jeon T, Ouyang M, Chalak L, Bi Y, Rollins N, Dong Q, Huang H. Early development of functional network segregation revealed by connectomic analysis of the preterm human brain. Cerebral Cortex. 2017a;27:1949–1963. doi: 10.1093/cercor/bhw038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cao M, Huang H, He Y. Developmental connectomics from infancy through early childhood. Trends in Neurosciences. 2017b;40:494–506. doi: 10.1016/j.tins.2017.06.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Chiron C, Raynaud C, Mazière B, Zilbovicius M, Laflamme L, Masure MC, Dulac O, Bourguignon M, Syrota A. Changes in regional cerebral blood flow during brain maturation in children and adolescents. The Journal of Nuclear Medicine: Official Publication, Society of Nuclear Medicine. 1992;33:696–703. [PubMed] [Google Scholar]
  12. Chiron C, Jambaque I, Nabbout R, Lounes R, Syrota A, Dulac O. The right brain hemisphere is dominant in human infants. Brain. 1997;120 (Pt 6):1057–1065. doi: 10.1093/brain/120.6.1057. [DOI] [PubMed] [Google Scholar]
  13. Chugani HT, Phelps ME. Maturational changes in cerebral function in infants determined by 18FDG positron emission tomography. Science. 1986;231:840–843. doi: 10.1126/science.3945811. [DOI] [PubMed] [Google Scholar]
  14. Chugani HT, Phelps ME, Mazziotta JC. Positron emission tomography study of human brain functional development. Annals of Neurology. 1987;22:487–497. doi: 10.1002/ana.410220408. [DOI] [PubMed] [Google Scholar]
  15. Cusack R, McCuaig O, Linke AC. Methodological challenges in the comparison of infant fMRI across age groups. Developmental Cognitive Neuroscience. 2018;33:194–205. doi: 10.1016/j.dcn.2017.11.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Deming WE. Statistical Adjutement of Data. Wiley; 1943. [Google Scholar]
  17. Detre JA, Alsop DC. Perfusion magnetic resonance imaging with continuous arterial spin labeling: methods and clinical applications in the central nervous system. European Journal of Radiology. 1999;30:115–124. doi: 10.1016/s0720-048x(99)00050-9. [DOI] [PubMed] [Google Scholar]
  18. Dittmer JC, Dawson RM. The isolation of a new lipid, triphosphoinositide, and monophosphoinositide from ox brain. The Biochemical Journal. 1961;81:535–540. doi: 10.1042/bj0810535. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Doria V, Beckmann CF, Arichi T, Merchant N, Groppo M, Turkheimer FE, Counsell SJ, Murgasova M, Aljabar P, Nunes RG, Larkman DJ, Rees G, Edwards AD. Emergence of resting state networks in the preterm human brain. PNAS. 2010;107:20015–20020. doi: 10.1073/pnas.1007921107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Doyle-Thomas KAR, Lee W, Foster NEV, Tryfon A, Ouimet T, Hyde KL, Evans AC, Lewis J, Zwaigenbaum L, Anagnostou E, NeuroDevNet ASD Imaging Group Atypical functional brain connectivity during rest in autism spectrum disorders. Annals of Neurology. 2015;77:866–876. doi: 10.1002/ana.24391. [DOI] [PubMed] [Google Scholar]
  21. Fox PT, Raichle ME. Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects. PNAS. 1986;83:1140–1144. doi: 10.1073/pnas.83.4.1140. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Fox MD, Snyder AZ, Vincent JL, Raichle ME. Intrinsic fluctuations within cortical systems account for intertrial variability in human behavior. Neuron. 2007;56:171–184. doi: 10.1016/j.neuron.2007.08.023. [DOI] [PubMed] [Google Scholar]
  23. Fransson P, Skiöld B, Horsch S, Nordell A, Blennow M, Lagercrantz H, Aden U. Resting-state networks in the infant brain. PNAS. 2007;104:15531–15536. doi: 10.1073/pnas.0704380104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Gao W, Zhu H, Giovanello KS, Smith JK, Shen D, Gilmore JH, Lin W. Evidence on the emergence of the brain’s default network from 2-week-old to 2-year-old healthy pediatric subjects. PNAS. 2009;106:6790–6795. doi: 10.1073/pnas.0811221106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Greicius MD, Krasnow B, Reiss AL, Menon V. Functional connectivity in the resting brain: a network analysis of the default mode hypothesis. PNAS. 2003;100:253–258. doi: 10.1073/pnas.0135058100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Greicius MD, Srivastava G, Reiss AL, Menon V. Default-mode network activity distinguishes alzheimer’s disease from healthy aging: evidence from functional MRI. PNAS. 2004;101:4637–4642. doi: 10.1073/pnas.0308627101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Greicius MD, Supekar K, Menon V, Dougherty RF. Resting-state functional connectivity reflects structural connectivity in the default mode network. Cerebral Cortex. 2009;19:72–78. doi: 10.1093/cercor/bhn059. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Gur RC, Ragland JD, Reivich M, Greenberg JH, Alavi A, Gur RE. Regional differences in the coupling between resting cerebral blood flow and metabolism may indicate action preparedness as a default state. Cerebral Cortex. 2009;19:375–382. doi: 10.1093/cercor/bhn087. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Haynes RL, Borenstein NS, Desilva TM, Folkerth RD, Liu LG, Volpe JJ, Kinney HC. Axonal development in the cerebral white matter of the human fetus and infant. The Journal of Comparative Neurology. 2005;484:156–167. doi: 10.1002/cne.20453. [DOI] [PubMed] [Google Scholar]
  30. Herscovitch P, Raichle ME. What is the correct value for the brain -- blood partition coefficient for water? Journal of Cerebral Blood Flow and Metabolism. 1985;5:65–69. doi: 10.1038/jcbfm.1985.9. [DOI] [PubMed] [Google Scholar]
  31. Huang H, Jeon T, Sedmak G, Pletikos M, Vasung L, Xu X, Yarowsky P, Richards LJ, Kostovic I, Sestan N, Mori S. Coupling diffusion imaging with histological and gene expression analysis to examine the dynamics of cortical areas across the fetal period of human brain development. Cerebral Cortex. 2013;23:2620–2631. doi: 10.1093/cercor/bhs241. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Huang H., Vasung L. Gaining insight of fetal brain development with diffusion MRI and histology. International Journal of Developmental Neuroscience. 2014;32:11–22. doi: 10.1016/j.ijdevneu.2013.06.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Huttenlocher PR, Dabholkar AS. Regional differences in synaptogenesis in human cerebral cortex. The Journal of Comparative Neurology. 1997;387:167–178. doi: 10.1002/(sici)1096-9861(19971020)387:2&#x0003c;167::aid-cne1&#x0003e;3.0.co;2-z. [DOI] [PubMed] [Google Scholar]
  34. Innocenti GM, Price DJ. Exuberance in the development of cortical networks. Nature Reviews. Neuroscience. 2005;6:955–965. doi: 10.1038/nrn1790. [DOI] [PubMed] [Google Scholar]
  35. Jain V, Duda J, Avants B, Giannetta M, Xie SX, Roberts T, Detre JA, Hurt H, Wehrli FW, Wang DJJ. Longitudinal reproducibility and accuracy of pseudo-continuous arterial spin-labeled perfusion MR imaging in typically developing children. Radiology. 2012;263:527–536. doi: 10.1148/radiol.12111509. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Kim HG, Lee JH, Choi JW, Han M, Gho SM, Moon Y. Multidelay arterial spin-labeling MRI in neonates and infants: cerebral perfusion changes during brain maturation. AJNR. American Journal of Neuroradiology. 2018;39:1912–1918. doi: 10.3174/ajnr.A5774. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Knickmeyer RC, Gouttard S, Kang C, Evans D, Wilber K, Smith JK, Hamer RM, Lin W, Gerig G, Gilmore JH. A structural MRI study of human brain development from birth to 2 years. The Journal of Neuroscience. 2008;28:12176–12182. doi: 10.1523/JNEUROSCI.3479-08.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Lemaître H, Augé P, Saitovitch A, Vinçon-Leite A, Tacchella JM, Fillon L, Calmon R, Dangouloff-Ros V, Lévy R, Grévent D, Brunelle F, Boddaert N, Zilbovicius M. Rest functional brain maturation during the first year of life. Cerebral Cortex. 2021;31:1776–1785. doi: 10.1093/cercor/bhaa325. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Li A, Gong L, Xu F. Brain-state-independent neural representation of peripheral stimulation in rat olfactory bulb. PNAS. 2011;108:5087–5092. doi: 10.1073/pnas.1013814108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Liang X, Zou Q, He Y, Yang Y. Coupling of functional connectivity and regional cerebral blood flow reveals a physiological basis for network hubs of the human brain. PNAS. 2013;110:1929–1934. doi: 10.1073/pnas.1214900110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Liu P, Chalak LF, Krishnamurthy LC, Mir I, Peng S, Huang H, Lu H. T1 and T2 values of human neonatal blood at 3 tesla: dependence on hematocrit, oxygenation, and temperature. Magnetic Resonance in Medicine. 2016;75:1730–1735. doi: 10.1002/mrm.25775. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Liu F, Duan Y, Peterson BS, Asllani I, Zelaya F, Lythgoe D, Kangarlu A. Resting state cerebral blood flow with arterial spin labeling MRI in developing human brains. European Journal of Paediatric Neurology. 2018;22:642–651. doi: 10.1016/j.ejpn.2018.03.003. [DOI] [PubMed] [Google Scholar]
  43. Liu P., Qi Y, Lin Z, Guo Q, Wang X, Lu H. Assessment of cerebral blood flow in neonates and infants: a phase-contrast MRI study. NeuroImage. 2019;185:926–933. doi: 10.1016/j.neuroimage.2018.03.020. [DOI] [PubMed] [Google Scholar]
  44. Lynch CJ, Uddin LQ, Supekar K, Khouzam A, Phillips J, Menon V. Default mode network in childhood autism: posteromedial cortex heterogeneity and relationship with social deficits. Biological Psychiatry. 2013;74:212–219. doi: 10.1016/j.biopsych.2012.12.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Miller DJ, Duka T, Stimpson CD, Schapiro SJ, Baze WB, McArthur MJ, Fobbs AJ, Sousa AMM, Sestan N, Wildman DE, Lipovich L, Kuzawa CW, Hof PR, Sherwood CC. Prolonged myelination in human neocortical evolution. PNAS. 2012;109:16480–16485. doi: 10.1073/pnas.1117943109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Ouyang M, Liu P, Jeon T, Chalak L, Heyne R, Rollins NK, Licht DJ, Detre JA, Roberts TPL, Lu H, Huang H. Heterogeneous increases of regional cerebral blood flow during preterm brain development: preliminary assessment with pseudo-continuous arterial spin labeled perfusion MRI. NeuroImage. 2017;147:233–242. doi: 10.1016/j.neuroimage.2016.12.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Ouyang M, Dubois J, Yu Q, Mukherjee P, Huang H. Delineation of early brain development from fetuses to infants with diffusion MRI and beyond. NeuroImage. 2019a;185:836–850. doi: 10.1016/j.neuroimage.2018.04.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Ouyang M, Jeon T, Sotiras A, Peng Q, Mishra V, Halovanic C, Chen M, Chalak L, Rollins N, Roberts TPL, Davatzikos C, Huang H. Differential cortical microstructural maturation in the preterm human brain with diffusion kurtosis and tensor imaging. PNAS. 2019b;116:4681–4688. doi: 10.1073/pnas.1812156116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Padmanabhan A, Lynch CJ, Schaer M, Menon V. The default mode network in autism. Biological Psychiatry. Cognitive Neuroscience and Neuroimaging. 2017;2:476–486. doi: 10.1016/j.bpsc.2017.04.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Paulson OB, Hasselbalch SG, Rostrup E, Knudsen GM, Pelligrino D. Cerebral blood flow response to functional activation. Journal of Cerebral Blood Flow and Metabolism. 2010;30:2–14. doi: 10.1038/jcbfm.2009.188. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Peng Q, Ouyang M, Wang J, Yu Q, Zhao C, Slinger M, Li H, Fan Y, Hong B, Huang H. Regularized-ncut: robust and homogeneous functional parcellation of neonate and adult brain networks. Artificial Intelligence in Medicine. 2020;106:101872. doi: 10.1016/j.artmed.2020.101872. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Pfefferbaum A, Mathalon DH, Sullivan EV, Rawles JM, Zipursky RB, Lim KO. A quantitative magnetic resonance imaging study of changes in brain morphology from infancy to late adulthood. Archives of Neurology. 1994;51:874–887. doi: 10.1001/archneur.1994.00540210046012. [DOI] [PubMed] [Google Scholar]
  53. Pontzer H, Yamada Y, Sagayama H, Ainslie PN, Andersen LF, Anderson LJ, Arab L, Baddou I, Bedu-Addo K, Blaak EE, Blanc S, Bonomi AG, Bouten CVC, Bovet P, Buchowski MS, Butte NF, Camps SG, Close GL, Cooper JA, Cooper R, Das SK, Dugas LR, Ekelund U, Entringer S, Forrester T, Fudge BW, Goris AH, Gurven M, Hambly C, El Hamdouchi A, Hoos MB, Hu S, Joonas N, Joosen AM, Katzmarzyk P, Kempen KP, Kimura M, Kraus WE, Kushner RF, Lambert EV, Leonard WR, Lessan N, Martin C, Medin AC, Meijer EP, Morehen JC, Morton JP, Neuhouser ML, Nicklas TA, Ojiambo RM, Pietiläinen KH, Pitsiladis YP, Plange-Rhule J, Plasqui G, Prentice RL, Rabinovich RA, Racette SB, Raichlen DA, Ravussin E, Reynolds RM, Roberts SB, Schuit AJ, Sjödin AM, Stice E, Urlacher SS, Valenti G, Van Etten LM, Van Mil EA, Wells JCK, Wilson G, Wood BM, Yanovski J, Yoshida T, Zhang X, Murphy-Alford AJ, Loechl C, Luke AH, Rood J, Schoeller DA, Westerterp KR, Wong WW, Speakman JR, IAEA DLW Database Consortium Daily energy expenditure through the human life course. Science. 2021;373:808–812. doi: 10.1126/science.abe5017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GL. A default mode of brain function. PNAS. 2001;98:676–682. doi: 10.1073/pnas.98.2.676. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Raichle ME. Two views of brain function. Trends in Cognitive Sciences. 2010;14:180–190. doi: 10.1016/j.tics.2010.01.008. [DOI] [PubMed] [Google Scholar]
  56. Raichle ME. The brain’s default mode network. Annual Review of Neuroscience. 2015;38:433–447. doi: 10.1146/annurev-neuro-071013-014030. [DOI] [PubMed] [Google Scholar]
  57. Rakic P. Radial versus tangential migration of neuronal clones in the developing cerebral cortex. PNAS. 1995;92:11323–11327. doi: 10.1073/pnas.92.25.11323. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Satterthwaite TD, Shinohara RT, Wolf DH, Hopson RD, Elliott MA, Vandekar SN, Ruparel K, Calkins ME, Roalf DR, Gennatas ED, Jackson C, Erus G, Prabhakaran K, Davatzikos C, Detre JA, Hakonarson H, Gur RC, Gur RE. Impact of puberty on the evolution of cerebral perfusion during adolescence. PNAS. 2014;111:8643–8648. doi: 10.1073/pnas.1400178111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Settergren G, Lindblad BS, Persson B. Cerebral blood flow and exchange of oxygen, glucose, ketone bodies, lactate, pyruvate and amino acids in infants. Acta Paediatrica Scandinavica. 1976;65:343–353. doi: 10.1111/j.1651-2227.1976.tb04896.x. [DOI] [PubMed] [Google Scholar]
  60. Sidman RL, Rakic P. Neuronal migration, with special reference to developing human brain: a review. Brain Research. 1973;62:1–35. doi: 10.1016/0006-8993(73)90617-3. [DOI] [PubMed] [Google Scholar]
  61. Sidman RL, Rakic P. Development of the Human Central Nervous System, in Histology and Histopathology of the Nervous System. Thomas, Springfield; 1982. [Google Scholar]
  62. Silbereis JC, Pochareddy S, Zhu Y, Li M, Sestan N. The cellular and molecular landscapes of the developing human central nervous system. Neuron. 2016;89:248–268. doi: 10.1016/j.neuron.2015.12.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Smith SM, Fox PT, Miller KL, Glahn DC, Fox PM, Mackay CE, Filippini N, Watkins KE, Toro R, Laird AR, Beckmann CF. Correspondence of the brain’s functional architecture during activation and rest. PNAS. 2009;106:13040–13045. doi: 10.1073/pnas.0905267106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Smyser CD, Inder TE, Shimony JS, Hill JE, Degnan AJ, Snyder AZ, Neil JJ. Longitudinal analysis of neural network development in preterm infants. Cerebral Cortex. 2010;20:2852–2862. doi: 10.1093/cercor/bhq035. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Suzuki C, Kosugi M, Magata Y. Conscious rat PET imaging with soft immobilization for quantitation of brain functions: comprehensive assessment of anesthesia effects on cerebral blood flow and metabolism. EJNMMI Research. 2021;11:1–11. doi: 10.1186/s13550-021-00787-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Sydnor VJ, Larsen B, Bassett DS, Alexander-Bloch A, Fair DA, Liston C, Mackey AP, Milham MP, Pines A, Roalf DR, Seidlitz J, Xu T, Raznahan A, Satterthwaite TD. Neurodevelopment of the association cortices: patterns, mechanisms, and implications for psychopathology. Neuron. 2021;109:2820–2846. doi: 10.1016/j.neuron.2021.06.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Tau GZ, Peterson BS. Normal development of brain circuits. Neuropsychopharmacology. 2010;35:147–168. doi: 10.1038/npp.2009.115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Vaishnavi SN, Vlassenko AG, Rundle MM, Snyder AZ, Mintun MA, Raichle ME. Regional aerobic glycolysis in the human brain. PNAS. 2010;107:17757–17762. doi: 10.1073/pnas.1010459107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Varela M, Hajnal JV, Petersen ET, Golay X, Merchant N, Larkman DJ. A method for rapid in vivo measurement of blood T1. NMR in Biomedicine. 2011;24:80–88. doi: 10.1002/nbm.1559. [DOI] [PubMed] [Google Scholar]
  70. Varela M, Petersen ET, Golay X, Hajnal JV. Cerebral blood flow measurements in infants using look-locker arterial spin labeling. Journal of Magnetic Resonance Imaging. 2015;41:1591–1600. doi: 10.1002/jmri.24716. [DOI] [PubMed] [Google Scholar]
  71. Wang Z, Fernández-Seara M, Alsop DC, Liu WC, Flax JF, Benasich AA, Detre JA. Assessment of functional development in normal infant brain using arterial spin labeled perfusion MRI. NeuroImage. 2008;39:973–978. doi: 10.1016/j.neuroimage.2007.09.045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Washington SD, Gordon EM, Brar J, Warburton S, Sawyer AT, Wolfe A, Mease-Ference ER, Girton L, Hailu A, Mbwana J, Gaillard WD, Kalbfleisch ML, VanMeter JW. Dysmaturation of the default mode network in autism. Human Brain Mapping. 2014;35:1284–1296. doi: 10.1002/hbm.22252. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Yakovlev P. Regional Development of the Brain in Early Life. Blackwell Scientific Publications; 1967. The myelogenetic cycles of regional maturation of the brain; pp. 3–70. [Google Scholar]
  74. Yu Q, Ouyang A, Chalak L, Jeon T, Chia J, Mishra V, Sivarajan M, Jackson G, Rollins N, Liu S, Huang H. Structural development of human fetal and preterm brain cortical plate based on population-averaged templates. Cerebral Cortex. 2016;26:4381–4391. doi: 10.1093/cercor/bhv201. [DOI] [PMC free article] [PubMed] [Google Scholar]

Editor's evaluation

Timothy E Behrens 1

In this paper, the authors find a link between the emergence of functional connectivity (FC) and changes in regional Cerebral Blood Flow (rCBF) in human infancy from birth to 24 months of age, which will be of interest to the increasing field investigating how the establishment of the brain's functional organization is linked to neurodevelopmental and psychiatric conditions. The data quality and complementarity are impressive for infants over this developmental period (0-2 years). Most of the key claims of the manuscript are well supported by the data. However, the relatively sparse sample and cross-sectional nature do limit interpretation.

Decision letter

Editor: Timothy E Behrens1
Reviewed by: Jessica Dubois2

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 "Infant brain regional cerebral blood flow dynamics supporting emergence of vital functional networks" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Timothy Behrens as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Jessica Dubois (Reviewer #1).

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:

The reviewers found much to like about this study, but also raised some concerns. I am including the individual reviews below, but I am going to highlight the ones that will be a particular concern to the reviewers on revision.

1) There is a general concern that the rCBF differences you see across brain regions (values that seem to follow a spatial gradient from inferior to superior and from posterior to anterior) might reflect a technical issue/bias in relation to anatomy and physiological changes, as opposed to specific maturational changes and differences across brain networks including the DMN. This is alluded to in several of the reviewer comments, particularly but not exclusively in R3. Several strategies are suggested for alleviating this concern.

2) The use of the same imaging parameters across the population might lead to an underestimation of rCBF in the younger infants. If there is evidence against this, it should be supplied.

3) Similarly using the older infants to get the ROIs may lead to a selection bias. R2 has suggestions for unbiased analysis strategies.

4) The manuscript sells itself as a description of connectivity, broadly defined, but focuses on DMN. All reviewers think this needs clear justification in the intro, and potentially changes to the title and abstract. The word "vital" should also be removed from the title.

Reviewer #1 (Recommendations for the authors):

– Regarding rCBF measures in DMN network, it would be interesting to evaluate correlations across DMN regions: this might highlight that the joint functioning of these regions requires similar changes in rCBF across ages, which would mean a kind of "physiological connectivity" throughout development.

– Evaluating possible rCBF asymmetries between left and right hemispheres would be interesting given previous observations of the two following studies (currently missing in the list of references):

Chiron C, Jambaque I, Nabbout R, Lounes R, Syrota A, Dulac O. The right brain hemisphere is dominant in human infants. Brain. 1997 Jun;120 ( Pt 6):1057-65. doi: 10.1093/brain/120.6.1057.

Lemaître H, Augé P, Saitovitch A, Vinçon-Leite A, Tacchella JM, Fillon L, Calmon R, Dangouloff-Ros V, Lévy R, Grévent D, Brunelle F, Boddaert N, Zilbovicius M. Rest Functional Brain Maturation during the First Year of Life. Cereb Cortex. 2021 Feb 5;31(3):1776-1785. doi: 10.1093/cercor/bhaa325.

Reviewer #2 (Recommendations for the authors):

1. Whilst I understand the focus of the study on putative functional networks which have been shown to mature at different rates, this approach does limit the study somewhat. The title and the first part of the introduction suggest that the work is about whole brain development, but then the abstract and paper suddenly focus on the default mode network. Whilst the reason for this focus becomes apparent with reading, I found the motivation for their approach therefore somewhat unclear/inconsistent.

2. Could the authors better describe in the mansucript why they chose to focus only on three networks rather than whole brain approaches or suggest what they might consider doing (such as linked ICA together with anatomical data) in future studies given the available data?

3. With the networks specifically, it is possible that the use of the DMN maps derived from the older infants only, may bias the results as it could then be argued that it is unsurprising that connectivity (and potentially rCBF is less) in those infants. Would it be possible to see if an age-specific representation of the network (using a method such as dual regression) would still yield the same results?

4. It is mentioned in the results that the infants were sedated for study – the type of sedation should be mentioned (the drug and dose) and the possible implications on the results (especially the CBF) considered. This is also the case for the clinical history of the infants – a large proportion of them had a history of seizures, which is a concern as it questions the generalisability of the results to the healthy brain (even if they are apparently neurologically normal).

5. As mentioned above, it is a concern that the same acquisition and processing parameters were used for all of the infants. The T1 of arterial blood for example, seems to have been derived from published neonatal values which are likely to differ greatly from those needed for a 2 year old infant. In contrast, arterial transit time is likely to vary across development, and so the use of the same post-label delay time of 1600ms (which is markedly shorter than the value of 2000ms suggested in the Alsop white paper) for all infants is a concern – as this may lead to an underestimation of CBF in the youngest infants. Partial voluming effects will be worse in the youngest infants (especially with a slice thickness of 5mm), which could also lead to an underestimation of CBF as separation of the white matter and cortex is challenging.

6. I find the suggestion that the rCBF measures can tell us about changes in brain metabolism rather speculative – as there are no other measures of oxygen extraction for example to specifically justify this. As the data is cross-sectional, mechanistic insight such as understanding whether one precedes the other is not possible. Figure 6 for example is not justified by the data, as it is simply not possible to know from their data if the suggested physiological changes are actually taking place.

7. The language in parts is a little unclear and therefore could benefit from clarification. Example include: line 53 "increasing feed of energy to fuel the brain development" or line 56 "impeding fundamental view of energy expenditure across functional systems of early developing brain"

Specific points:

1. I would prefer rewording of the title if possible – the use of the word "vital" does not seem appropriate, as this is not demonstrated in the work.

2. In the first line of the abstract, the sentence reads rapid cerebral blood flow (rCBF) which incorrectly implies that this is what the acronym stands for.

3. The abstract mentions "rCBF dynamics" – which I would caution against. In this work, they are really only looking at the mean rCBF across the acquisition, rather than its specific dynamics across the volumes acquired.

4. Suggest rewording of line 30 "We found faster rCBF increases in the DMN than other regions." as whilst this is true in comparison to the visual/SM networks – it wasn't demonstrated in comparison to all other regions?

5. The statement in line 122 of "high quality of the rCBF maps" is not justified as its not clear how this "quality" has been defined – suggest rephrasing, please.

6. The values of rCBF on line 141 should state they are the mean across all of the infants and should include some estimate of the variance.

7. I note that rCBF is consistently higher at the base of the brain and occipital regions – is this a recognized pattern in the literature? If not, could it be because vascular crushers were not used and there is overestimation due to the residual blood in the basilar artery?

Reviewer #3 (Recommendations for the authors):

The manuscript describes a study of 48 infants aged 0-24 months who underwent resting state functional MRI (rs-fcMRI) and arterial spin labeling regional blood flow (rCBF) measurements. The authors focused on the default mode network (DMN), visual cortex (VIS), and somatomotor cortex (SM). They found faster increases in rCBF in the DMN with strongly coupled increases of rCBF to network "strength" measured through rs-fcMRI. They concluded that the faster increase in rCBF in the DMN takes place to meet the metabolic demands of DMN maturation.

The authors found stronger correlations between increases in rCBF and rs-fcMRI in the DMN than in VIS and SM. They suggest that this indicates that changes in rCBF constitute a physiological underpinning of DMN maturation. However, it is not clear that the underlying assertion that increases in rs-fcMRI (developmental maturation) and increases in rCBF are tightly linked is supported by the data.

First, VIS and SM are both more mature than DMN at birth, yet VIS has the highest rCBF and SM has the lowest. If rCBF and maturity were tightly coupled, one would expect both to have relatively high rCBF values.

Second, the data shown in Figure 2A show that while rCBF increases with maturation, it does so along posterior-to-anterior and inferior-to-superior gradients which do not appear to correlate with regional variability in maturation. The SM region chosen for the study is located mainly in the leg area, which is primarily in the superior and medial SM cortex. The more inferior SM areas, corresponding to the arm and face, are not included. Given an inferior-to-superior gradient in the maturational increase in rCBF, the more superior portion of SM chosen is relatively late to show an increase in rCBF.

Third, there are fairly large differences in the rCBF of subregions of the DMN, though one would expect them to be similar if they are maturing as a unit.

Finally, the stronger correlation between increases in rCBF and rs-fcMRI metrics for the DMN may reflect its greater increase in rCBF during the time period studied. This may be due to the finding that VIS, located posteriorly and inferiorly, has already undergone its greatest increase in rCBF; while SM, located superiorly, has yet to undergo its greatest increase in rCBF. Thus both have relatively modest changes in rCBF during the period studied. The steeper rise in rCBF in the DMN, which may be by virtue of its location, would allow for a stronger correlation with maturational increases in rs-fcMRI parameters.

Since the gradient in increasing rCBF during development appears to be based on relative location, it is important to consider whether the superior-inferior gradient might be a technical artifact related to applying labeling pulses to the neck, which is getting longer, or changes in flow velocities with age.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Infant brain regional cerebral blood flow increases supporting emergence of the default-mode network" for further consideration by eLife. Your revised article has been evaluated by Timothy Behrens (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

As you will see the reviewers have some concerns remaining. The largest concern is that measurement differences in acquisition might have introduced bias. It would clearly be best if you could provide some evidence, or even simulation, to address these points (detailed below), but if this is not possible then please explicitly acknowledge these concerns in the discussion and discuss the possible effects of this bias on your results.

Reviewer #1 (Recommendations for the authors):

The authors have partly responded to the reviewers' comments in my opinion. Some points still seem to me to be questionable in the manuscript, whereas some answers provided by the authors in their replies would have benefited from being added to the manuscript.

– Essential revisions 1 and 2: unless I am mistaken, these elements have not been added to the manuscript. Perhaps this would be relevant for readers who don't read the reviewers' comments and the authors' answers in addition to the article.

– R1-1: The answer provided might seem insufficient. The grey matter masks shown seem to me to clearly contain white matter, at least in the youngest infants for whom the cortex is very thin. The problem of reliably identifying the cortex to limit partial volume effects is definitely complex, and pCASL acquisition with thick slices doesn't avoid this problem anyway. That said, I still trust the measures performed, but I think it is important to discuss this point so that the reader is aware of this potential measurement bias. One way of doing this would be to estimate what the bias might be as a function of age (based on the estimated proportions of grey/white matter in the masks considered, and estimated rCBF in each tissue).

– R1-2: Again, the answer provided seems incomplete. When two measures depend on the same variable (age), their correlation can be partly driven by the common correlation, and it cannot then be considered as specific. The similarity between figures 3b and 5d on the lateral side draws attention to this point. Unless I am mistaken, the measures of functional connectivity in the visual and sensorimotor networks are much less age-dependent than in the DMN network (Figure 1-sup2), whereas the rCBF is age-dependent in all regions: this difference might drive the difference in FC-rCBF correlation observed across networks. A simple way to "prove" that the FC-rCBF correlation is indeed specific for the DMN would be to perform all these correlation analyses after removing all age effects in all measurements (region-wise or vertex-wise). These results would be convincing.

– R1-4: it seems to me that the order of the method paragraphs should be changed to follow the new order of the results.

– R1-11: citation of the Cusack et al. paper would be welcome in the manuscript.

– Introduction p3: the sentence "there have been no known whole-brain mappings of heterogeneous infant brain regional cerebral blood flow (rCBF) changes across landmark infant ages thus far" is incorrect (see the Lemaitre et al. paper pointed out in the previous review).

Reviewer #2 (Recommendations for the authors):

Thank you for the detailed answers to the questions and amendments to the manuscript. Whilst I agree with some of their responses, I feel there are still some unresolved issues which I think are fundamental to resolve:

1. The use of similar acquisition parameters for all age groups: as far I understand it, the method proposed by Aslan et al. using PCA to calculate the labelling efficiency still requires the T1 of arterial blood as a parameter – for which neonatal values (according to their references) have been used. It would be good to know what would happen if more developmentally appropriate values were used in their CBF estimations. Secondly, the Aslan method provides a single corrective factor (and estimate of the labelling efficiency) for the CBF estimation based on the PCA measurement (which is global). Therefore it cannot correct for any regional effects of using a different post-label delay in the ASL acquisition – with knock-on effects for the regional CBF estimation (such as the spatial bias leading to the inferior-superior gradient for example). Would it be possible for the authors to provide some empirical evidence that the post-label delay has not influenced their results?

2. With regards to the inferior-superior gradient, the references that the authors cite do not show the same kind of effects. Whilst many of them do suggest that CBF is higher posteriorly and is higher in the occipital lobe – it is not a "hot spot" in the same way as seen in the maps here (especially at the youngest ages). The maps in the Lamaitre work for example look very different (see figure 2 in their paper) to those shown in figure 3 here.

3. I am unconvinced by the argument that calculating grey matter maps in native high-resolution space and registering them together with the low-resolution CBF map into standard space thresholded probabilistic maps resolve the partial voluming issue. Fundamentally this does not change the base acquisition voxel which is very large and is the same across all of the infants studied – so differences relative to the size of the brain/cortex and partial voluming in the acquisition itself in the acquisition voxel are surely still an issue?

eLife. 2023 Jan 24;12:e78397. doi: 10.7554/eLife.78397.sa2

Author response

Essential revisions:

The reviewers found much to like about this study, but also raised some concerns. I am including the individual reviews below, but I am going to highlight the ones that will be a particular concern to the reviewers on revision.

1) There is a general concern that the rCBF differences you see across brain regions (values that seem to follow a spatial gradient from inferior to superior and from posterior to anterior) might reflect a technical issue/bias in relation to anatomy and physiological changes, as opposed to specific maturational changes and differences across brain networks including the DMN. This is alluded to in several of the reviewer comments, particularly but not exclusively in R3. Several strategies are suggested for alleviating this concern.

We thank the reviewers for this comment. When initially inspecting the data, we too suspect the spatial gradient might reflect a technical bias in relation to anatomy and physiological changes. We carefully reviewed data before the original submission. We respectively disagree that rCBF difference across brain regions reflects a technical issue/bias due to following justification. The pattern of rCBF spatial gradient from inferior to superior and from posterior to anterior has also been consistently reported in various studies across different age groups. For example, previous studies have observed higher rCBF values at the base of the brain and occipital regions than in the frontal regions in infant and neonate brains (e.g., Kim et al., 2018; Lemaître et al., 2021; Wang et al., 2008). This spatial gradient pattern has also been reported in Satterthwaite’s study with rCBF results of 922 youths aged 8-22 years (Satterthwaite et al., 2014). Consistency across platform and across studies would only indicate bias of arterial spin labelling (ASL) perfusion MRI technique in general. However, ASL perfusion MRI is a noninvasive quantitative method that has been validated extensively against “gold standard” 15O-PET as well as across platforms, including it reproducibility. Although future investigation of rCBF measurements from infants of a similar age range with a larger sample size is warranted, existing literature indicates high rCBF values at the base of the brain and occipital regions are unlikely due to the residual blood in the basilar artery.

Kim, H.G., Lee, J.H., Choi, J.W., Han, M., Gho, S.M., and Moon, Y., (2018). Multidelay arterial spin-labeling MRI in neonates and infants: cerebral perfusion changes during brain maturation. American Journal of Neuroradiology 39(10): 1912-1918.

Lemaître, H., Augé, P., Saitovitch, A., Vinçon-Leite, A., Tacchella, J.M., Fillon, L., Calmon, R., Dangouloff-Ros, V., Lévy, R., Grévent, D., et al. (2021). Rest functional brain maturation during the first year of life. Cereb. Cortex 31, 1776-1785.

Satterthwaite, T.D., Shinohara, R.T., Wolf, D.H., Hopson, R.D., Elliott, M.A., Vandekar, S.N., Ruparel, K., Calkins, M.E., Roalf, D.R., Gennatas, E.D., et al. (2014). Impact of puberty on the evolution of cerebral perfusion during adolescence. Proc. Natl. Acad. Sci. USA 111, 8643-8648.

Wang, Z., Fernández-Seara, M., Alsop, D.C., Liu, W.C., Flax, J.F., Benasich, A.A., and Detre, J.A., (2008). Assessment of functional development in normal infant brain using arterial spin labeled perfusion MRI. Neuroimage 39(3), 973-978.

2) The use of the same imaging parameters across the population might lead to an underestimation of rCBF in the younger infants. If there is evidence against this, it should be supplied.

The reviewer’s comment is valid. We would like to address the comment in the following aspects: (1) infant physiological parameters including PLD and imaging parameter selections are in the interval between neonates and children, and probably closer to children. For example, according to the Alsop white paper, a PLD of 2000ms was recommended for neonates, while a PLD of 1500ms was recommended for children. In our cohort, a PLD time of 1600ms was tailored for the studied infant age range. (2) Extra phase contrast MRI was acquired for all infants. And all rCBF maps of individual infants obtained with ASL scans in our study were calibrated using phase contrast MRI by applying a normalization factor so that there should not be any underestimation of rCBF in the younger infants. As suggested by Aslan et al. (Aslan et al., 2010), this approach largely minimizes the effects of individual variations in processing parameters such as T1 of arterial blood or labeling efficiency in the final estimation of rCBF measurement. Please also see our response to R2-5 for details.

Aslan, S., Xu, F., Wang, P.L., Uh, J., Yezhuvath, U.S., Van Osch, M., and Lu, H. (2010). Estimation of labeling efficiency in pseudocontinuous arterial spin labeling. Magn Reson Med. 63, 765-771.

3) Similarly using the older infants to get the ROIs may lead to a selection bias. R2 has suggestions for unbiased analysis strategies.

We thank the reviewer for this comment. The reviewer raised a very good point. We have strong counter example showing consistent network ROIs generated from older infant data do not lead to biased higher functional connectivity in the older infants, as shown in nonsignificant age-related change of functional connectivity within the visual (Vis) and somatosensory (SM) networks in Figure 1—figure supplement 2.

We had carefully evaluated approach of consistent ROIs used in the manuscript and approach of age-specific ROIs before adopting consistent ROI approach before original submission. Age-specific ROIs may lead to inconsistent measurement and add artificial bias to the measurement aimed for quantifying age-related blood flow changes. Specifically, if different and age-specific ROIs were applied, it would be difficult to delineate if the blood flow change is due to age-related brain maturation or ROI change. Secondly, DMN is not reproducible and reliable at the younger infants characterized by weak functional connections between PCC and other DMN regions (Alcauter et al., 2014). It is not feasible to establish age-specific DMN ROIs for younger infants when the ROIs cannot be reliably identified. And including younger infants to generate ROIs adds extra noise resulting in inaccurate ROIs.

Alcauter, S., Lin, W., Smith, J.K., Short, S.J., Goldman, B.D., Reznick, J.S., Gilmore, J.H., and Gao, W. (2014). Development of thalamocortical connectivity during infancy and its cognitive correlations. Journal of Neuroscience 34(27), 9067-9075.

4) The manuscript sells itself as a description of connectivity, broadly defined, but focuses on DMN. All reviewers think this needs clear justification in the intro, and potentially changes to the title and abstract. The word "vital" should also be removed from the title.

We thank the reviewers for this comment. We used broadly defined network as other network (visual and somatosensory) results were also presented in the manuscript. But we agree those networks were only used as reference networks for focused DMN. Therefore, following suggestions, we have renamed title as “Infant brain regional cerebral flow increases supporting emergence of the default-mode network” and modified abstract and manuscript accordingly too. Please also see our responses to R1-5 and R2-8.

Reviewer #1 (Recommendations for the authors):

– Regarding rCBF measures in DMN network, it would be interesting to evaluate correlations across DMN regions: this might highlight that the joint functioning of these regions requires similar changes in rCBF across ages, which would mean a kind of "physiological connectivity" throughout development.

– Evaluating possible rCBF asymmetries between left and right hemispheres would be interesting given previous observations of the two following studies (currently missing in the list of references):

Chiron C, Jambaque I, Nabbout R, Lounes R, Syrota A, Dulac O. The right brain hemisphere is dominant in human infants. Brain. 1997 Jun;120 ( Pt 6):1057-65. doi: 10.1093/brain/120.6.1057.

Lemaître H, Augé P, Saitovitch A, Vinçon-Leite A, Tacchella JM, Fillon L, Calmon R, Dangouloff-Ros V, Lévy R, Grévent D, Brunelle F, Boddaert N, Zilbovicius M. Rest Functional Brain Maturation during the First Year of Life. Cereb Cortex. 2021 Feb 5;31(3):1776-1785. doi: 10.1093/cercor/bhaa325.

We thank the reviewer for this comment. To reveal sort of “physiological connectivity”, the region-wise rCBF correlation was conducted across infants, we found that all the correlations were significant (p<0.001, FDR corrected), with higher correlations between the DMN subregions (i.e. DMN_PCC, DMN_MPFC, DMN_IPL, DMN_ITC), as shown in Author response image 1.

Author response image 1. Physiological connectivity.

Author response image 1.

Open in a new tab

To evaluate possible rCBF asymmetry, we compared rCBF in the same brain network regions between two hemispheres, with the results shown in Author response image 3. Significantly higher rCBF (t=3.82, p <0.05) was found in the SM network regions in the right hemisphere (50.8 ± 1.67 ml/100g/min), compared to that in the left hemisphere (47.8 ± 1.43 ml/100g/min). The finding of rCBF asymmetry is consistent with the previous study (Chiron et al., 1997; Lemaître et al., 2021) and has been added to the Faster rCBF increases in the DMN hub regions during infant brain development subsection in RESULT section.

Author response image 3. The gray matter mask and segmented cortical mask overlaid on rCBF map of three representative infants aged 3, 6, and 20 months in the template space.

Author response image 3.

Open in a new tab

The gray matter mask on the left panel was created to minimize the contamination of white matter and CSF while keeping the continuity of the cortical gray matter mask across the cerebral cortex. The contour of the gray matter mask was highlighted with bule line.

Chiron, C., Jambaque, I., Nabbout, R., Lounes, R., Syrota, A., and Dulac, O. (1997). The right brain hemisphere is dominant in human infants. Brain: a journal of neurology, 120, 1057-1065.

Lemaître, H., Augé, P., Saitovitch, A., Vinçon-Leite, A., Tacchella, J.M., Fillon, L., Calmon, R., Dangouloff-Ros, V., Lévy, R., Grévent, D., et al. (2021). Rest functional brain maturation during the first year of life. Cereb. Cortex 31, 1776-1785.

Author response image 2. Significant higher rCBF in the sensorimotor (SM) network regions in right hemisphere.

Author response image 2.

Open in a new tab

Reviewer #2 (Recommendations for the authors):

1. Whilst I understand the focus of the study on putative functional networks which have been shown to mature at different rates, this approach does limit the study somewhat. The title and the first part of the introduction suggest that the work is about whole brain development, but then the abstract and paper suddenly focus on the default mode network. Whilst the reason for this focus becomes apparent with reading, I found the motivation for their approach therefore somewhat unclear/inconsistent.

We agree with the reviewer. The background of whole brain development and measurement of other networks are indeed to provide reference and to emphasize the extraordinarily rCBF development of the default mode network during infancy. To make the focus of the default mode network clear for the entire manuscript, we have changed manuscript title from “Infant brain regional cerebral blood flow increases supporting emergence of vital functional networks” to “Infant brain regional cerebral blood flow increases supporting emergence of the default-mode network” to emphasize the default-mode network. Please also see our response to the Essential Revisions comment #4 summarized by the editor.

2. Could the authors better describe in the mansucript why they chose to focus only on three networks rather than whole brain approaches or suggest what they might consider doing (such as linked ICA together with anatomical data) in future studies given the available data?

We thank the reviewer for this comment. These three networks have been relatively more extensively studied in the literatures (Gilmore et al., 2018; Smith et al., 2009). As the initial effort to tackle the relationship between rCBF and functional networks, we only incorporated clearly defined and extensively studied infant networks, which are the three networks described in this paper. Future studies of other infant networks including language network are warranted. In fact, we already conducted preliminary analysis investigating relationship between rCBF and functional connectivity in the language network (Ouyang et al., 2020).

Gilmore, J.H., Knickmeyer, R.C., and Gao, W. (2018). Imaging structural and functional brain development in early childhood. Nat. Rev. Neurosci. 19(3), 123-137.

Ouyang, M., Yu, Q., Kang, H., Peng, Y., Hong, B., and Huang, H. (2020). Delineation of language network maturation during infancy with multi-modal perfusion and functional MRI. Proceedings of ISMRM (Magna Cum Laude Award).

Smith, S.M., Fox, P.T., Miller, K.L., Glahn, D.C., Fox, P.M., Mackay, C.E., Filippini, N., Watkins, K.E., Toro, R., Laird, A.R., et al. (2009). Correspondence of the brain's functional architecture during activation and rest. Proc. Natl. Acad. Sci. USA 106, 13040-13045.

3. With the networks specifically, it is possible that the use of the DMN maps derived from the older infants only, may bias the results as it could then be argued that it is unsurprising that connectivity (and potentially rCBF is less) in those infants. Would it be possible to see if an age-specific representation of the network (using a method such as dual regression) would still yield the same results?

This comment was included as Essential Revisions comment #3 summarized by the editor. Please see our response to that comment. Briefly, we argued that age-specific ROIs may lead to inconsistent measurement and add artificial bias to the measurement aimed for quantifying age-related blood flow changes. We also have strong counter example showing consistent network ROIs generated from older infant data do not lead to biased higher functional connectivity in the older infants, as shown in nonsignificant age-related change of functional connectivity within the visual (Vis) and somatosensory (SM) networks in Figure 1—figure supplement 2.

4. It is mentioned in the results that the infants were sedated for study – the type of sedation should be mentioned (the drug and dose) and the possible implications on the results (especially the CBF) considered. This is also the case for the clinical history of the infants – a large proportion of them had a history of seizures, which is a concern as it questions the generalisability of the results to the healthy brain (even if they are apparently neurologically normal).

We thank the reviewer for this comment. Chloral hydrate, with a dose of 0.5 ml/kg and no more than 10 ml in total, was taken orally for each infant before scanning for sedation. Previous studies (Li et al., 2011; Suzuki et al., 2021) suggested no significant impact of chloral hydrae on CBF or sensory function. The description has been added to the Data acquisition subsection in Materials and Methods section.

We apologize that we stated history of seizures in error. Most infants had a simple febrile convulsion, rather than seizures with fever. These infants had no abnormality in the further neurological evaluation. Occasional simple febrile convulsion will not cause long-term brain function damage, as shown in the literature (Patterson et al., 2013; Sawires et al., 2021; Verity et al.,1992;). This statement has been corrected in the Data acquisition subsection in Materials and Methods section.

Li, A., Gong, L., and Xu, F. (2011). Brain-state–independent neural representation of peripheral stimulation in rat olfactory bulb. Proc. Natl. Acad. Sci. USA 108(12), 5087-5092.

Patterson, J.L., Carapetian, S.A., Hageman, J.R., and Kelley, K.R. (2013). Febrile seizures. Pediatric annals 42(12), 258-263.

Sawires, R., Buttery, J., and Fahey, M. (2021). A Review of Febrile Seizures: Recent Advances in Understanding of Febrile Seizure Pathophysiology and Commonly Implicated Viral Triggers. Frontiers in Pediatrics, 9.

Suzuki, C., Kosugi, M., and Magata, Y. (2021). Conscious rat PET imaging with soft immobilization for quantitation of brain functions: comprehensive assessment of anesthesia effects on cerebral blood flow and metabolism. EJNMMI research 11(1), 1-11.

Verity, C.M., and Golding, J. (1991). Risk of epilepsy after febrile convulsions: a national cohort study. British Medical Journal 303(6814), 1373-1376.

5. As mentioned above, it is a concern that the same acquisition and processing parameters were used for all of the infants. The T1 of arterial blood for example, seems to have been derived from published neonatal values which are likely to differ greatly from those needed for a 2 year old infant. In contrast, arterial transit time is likely to vary across development, and so the use of the same post-label delay time of 1600ms (which is markedly shorter than the value of 2000ms suggested in the Alsop white paper) for all infants is a concern – as this may lead to an underestimation of CBF in the youngest infants. Partial voluming effects will be worse in the youngest infants (especially with a slice thickness of 5mm), which could also lead to an underestimation of CBF as separation of the white matter and cortex is challenging.

We thank the reviewer for bringing up this good point. This comment has been partly included as the Essential Revisions comment #2 by the editor. Here, we would like to address the comment more systematically in the following aspects below: (1) infant physiological parameters including arterial transit times (ATT) and imaging parameter selections in the interval between neonates and adults, (2) calibration of rCBF maps with phase-contrast MRI, and (3) amelioration of partial volume effects with carefully testing multiple thresholds in the averaged gray matter probability maps.

In terms of acquisition, we used the same parameters on purpose for data harmonization. This was also a feasible way and less error-prone for MR technologists to handle the scan whenever there was infant getting recruited. As the reviewer pointed out, a post-label delay (PLD) time of 2000ms was recommended for neonates, while a PLD of 1500ms was recommended for children in the Alsop white paper. However, please note brain perfusion parameters of infants (1.37 to 24.36 months) in this study are quite different from those of neonates (usually less than 1 week or 0.25 month of age), as artery blood flow velocities increased rapidly during the first 6 months (Liu et al., 2019). Selected PLD should be larger than ATT. Mean ATT of infants were reported less than 1500ms (Varela et al., 2014) while median ATT of neonates was reported 2260ms (Kim et al. 2018). We therefore consider it valid to select PLD of 1600ms for the studied infants.

Regarding the processing parameters, we would like to point out extra phase contrast MRI was acquired and all rCBF maps of individual infants in our study have been calibrated with phase contrast whole brain cerebral blood flow by using a normalization factor. This approach will largely minimize the effect of individual variations in processing parameters such as T1 of arterial blood or labeling efficiency in the final estimation of rCBF measurement (Aslan et al., 2010).

In this study, to minimize the partial volume effect, individual rCBF map was generated in the individual space and calibrated by phase contrast MRI to minimize the individual variations of processing parameters such as T1 of arterial blood (Aslan et al., 2010). Cortical segmentation was also conducted in individual space. Then different types of images including rCBF map and gray matter segmentation probability map in the individual space were normalized into the template space. An averaged gray matter probability map was generated after inter-subject normalization. After carefully testing multiple thresholds in the averaged gray matter probability maps, 40% probability minimizing the contamination of white matter and CSF while keeping the continuity of the cortical gray matter mask across the cerebral cortex was used to generate the binary gray matter mask shown on the left panel of Author response image 3. As demonstrated in the right three panels in Author response image 3, the rCBF measure in the cortical mask in the template space is consistent across ages for accurate and reliable voxelwise comparison across age.

Aslan, S., Xu, F., Wang, P.L., Uh, J., Yezhuvath, U.S., Van Osch, M., and Lu, H., (2010). Estimation of labeling efficiency in pseudocontinuous arterial spin labeling. Magnetic resonance in medicine 63(3), 765-771.

Kim, H.G., Lee, J.H., Choi, J.W., Han, M., Gho, S.M., and Moon, Y., (2018). Multidelay arterial spin-labeling MRI in neonates and infants: cerebral perfusion changes during brain maturation. American Journal of Neuroradiology 39(10), 1912-1918.

Liu, P., Qi, Y., Lin, Z., Guo, Q., Wang, X., and Lu, H., (2019). Assessment of cerebral blood flow in neonates and infants: a phase-contrast MRI study. Neuroimage 185, 926-933.

Varela, M., Petersen, E.T., Golay, X., and Hajnal, J.V., (2015). Cerebral blood flow measurements in infants using look–locker arterial spin labeling. Journal of Magnetic Resonance Imaging 41(6), 1591-1600.

6. I find the suggestion that the rCBF measures can tell us about changes in brain metabolism rather speculative – as there are no other measures of oxygen extraction for example to specifically justify this. As the data is cross-sectional, mechanistic insight such as understanding whether one precedes the other is not possible. Figure 6 for example is not justified by the data, as it is simply not possible to know from their data if the suggested physiological changes are actually taking place.

We agree with the reviewer that we should downplay previous Figure 6. Although we have made it clear it is our hypothesis that cerebral blood flow increase is associated with synaptogenesis and synaptic efficacy increase, to downplay this hypothesis that may need further justification by the data, this previous standing-alone Figure 6 has been downgraded as a supplemental figure (current Figure 5—figure supplement 3).

7. The language in parts is a little unclear and therefore could benefit from clarification. Example include: line 53 "increasing feed of energy to fuel the brain development" or line 56 "impeding fundamental view of energy expenditure across functional systems of early developing brain"

We thank the reviewer for this comment. In the first paragraph of INTRODUCTION section, the words ‘increasing feed of energy to fuel the brain development’ has been changed to ‘increasing energy consumption of the brain’. The words ‘impeding fundamental view of energy expenditure across functional systems of early developing brain’ have been changed to ‘impeding understanding of energy expenditure across functional systems of early developing brain’.

Specific points:

1. I would prefer rewording of the title if possible – the use of the word "vital" does not seem appropriate, as this is not demonstrated in the work.

As suggested, the word ‘vital’ has been removed from the title and abstract.

2. In the first line of the abstract, the sentence reads rapid cerebral blood flow (rCBF) which incorrectly implies that this is what the acronym stands for.

We thank the reviewer for catching this typo. ‘most rapid cerebral blood flow (rCBF)’ has been revised to ‘most rapid regional cerebral blood flow (rCBF)’ in the abstract.

3. The abstract mentions "rCBF dynamics" – which I would caution against. In this work, they are really only looking at the mean rCBF across the acquisition, rather than its specific dynamics across the volumes acquired.

‘rCBF dynamics’ has been changed to ‘rCBF increase’ in the abstract. We have also changed ‘dynamics’ to ‘increases’ or ‘changes’ in the title and across the text to be consistent.

4. Suggest rewording of line 30 "We found faster rCBF increases in the DMN than other regions." as whilst this is true in comparison to the visual/SM networks – it wasn't demonstrated in comparison to all other regions?

As suggested, ‘other regions’ has been replaced by more specific ‘visual and sensorimotor networks’ in the abstract.

5. The statement in line 122 of "high quality of the rCBF maps" is not justified as its not clear how this "quality" has been defined – suggest rephrasing, please.

We thank the reviewer for this comment. ‘High quality of the rCBF maps’ has been rephased as ‘The rCBF maps with high gray/white matter contrasts’.

6. The values of rCBF on line 141 should state they are the mean across all of the infants and should include some estimate of the variance.

We thank the reviewer for this comment. Standard errors have been added in the subsection Faster rCBF increases in the DMN hub regions during infant brain development in the Results section to clarify the mean and stand errors in those rCBF measurements.

7. I note that rCBF is consistently higher at the base of the brain and occipital regions – is this a recognized pattern in the literature? If not, could it be because vascular crushers were not used and there is overestimation due to the residual blood in the basilar artery?

We thank the reviewer for bringing up this good point. This is a recognized pattern in the literature. This comment has been included as the Essential Revisions comment #1 by the editor. Please see our response to that comment. Relevant literature is also listed in the response to Essential Revisions comment #1.

Reviewer #3 (Recommendations for the authors):

The manuscript describes a study of 48 infants aged 0-24 months who underwent resting state functional MRI (rs-fcMRI) and arterial spin labeling regional blood flow (rCBF) measurements. The authors focused on the default mode network (DMN), visual cortex (VIS), and somatomotor cortex (SM). They found faster increases in rCBF in the DMN with strongly coupled increases of rCBF to network "strength" measured through rs-fcMRI. They concluded that the faster increase in rCBF in the DMN takes place to meet the metabolic demands of DMN maturation.

The authors found stronger correlations between increases in rCBF and rs-fcMRI in the DMN than in VIS and SM. They suggest that this indicates that changes in rCBF constitute a physiological underpinning of DMN maturation. However, it is not clear that the underlying assertion that increases in rs-fcMRI (developmental maturation) and increases in rCBF are tightly linked is supported by the data.

First, VIS and SM are both more mature than DMN at birth, yet VIS has the highest rCBF and SM has the lowest. If rCBF and maturity were tightly coupled, one would expect both to have relatively high rCBF values.

We thank the reviewer for this comment. We would like to clarify the finding is rCBF increase in a developmental period (not rCBF at baseline) is tightly coupled with maturity. Such rCBF increase is associated with the increase of energy consumption during brain development. So our finding is not contradictory to the observation that Vis has the highest rCBF and SM has the lowest. We would also note this observation of the rCBF distribution pattern of relatively higher rCBF in the VIS regions and lower rCBF int eh SM regions at a certain infant time point is consistent to previous PET study (Chugani et al., 1987) and ASL study (Wang et al., 2008).

Chugani, H.T., Phelps, M.E., and Mazziotta, J. C. (1987). Positron emission tomography study of human brain functional development. Annals of neurology 22(4), 487-497.

Wang, Z., Fernández-Seara, M., Alsop, D.C., Liu, W.C., Flax, J.F., Benasich, A.A., and Detre, J.A., (2008). Assessment of functional development in normal infant brain using arterial spin labeled perfusion MRI. Neuroimage 39(3), 973-978.

Second, the data shown in Figure 2A show that while rCBF increases with maturation, it does so along posterior-to-anterior and inferior-to-superior gradients which do not appear to correlate with regional variability in maturation. The SM region chosen for the study is located mainly in the leg area, which is primarily in the superior and medial SM cortex. The more inferior SM areas, corresponding to the arm and face, are not included. Given an inferior-to-superior gradient in the maturational increase in rCBF, the more superior portion of SM chosen is relatively late to show an increase in rCBF.

We thank the reviewer for this comment. We believe the maturation pattern is more complicated than relatively simplified posterior-to-anterior and inferior-to-superior gradients. It also presents a primary-to-association gradient reproducibly found in the literature (see Sydnor et al., 2021 for review). Such mixed patterns are consistent with the maturation pattern demonstrated in current Figure 3a (previous Figure 2a). More clear maturation gradient could benefit from future rCBF maps of higher signal-to-noise ratio and higher resolution.

The reviewer brought up an interesting point of incomplete SM areas. As stated in the manuscript, all ROIs including SM areas were data driven instead of mapped from SM from an atlas. Thus lack a portion of the functional area could inevitably happen. Using independent component analysis (ICA), we did find another ROI that might be related to inferior SM region, as shown in Author response image 4a. However, this ROI clearly included auditory cortex. To avoid confounding factor from other functional areas, this ROI was not included as the SM region. In Author response image 4b, we demonstrated although this ROI is located inferiorly, the maturation pattern of this ROI in terms of rCBF increase rate is not significantly different (p>0.05) from that of the SM included in the manuscript (Figure 1—figure supplement 1b) and located more superiorly.

Author response image 4. (a) An ROI revealed by ICA analysis included inferior SM and auditory cortex.

Author response image 4.

Open in a new tab

(b) In ROI shown in a, no significant difference (p > 0.05) of rCBF increase rate between this ROI and SM network ROI included in the manuscript was found. The rCBF of this ROI also increases significantly with age (r = 0.62, p<10-4).

Sydnor, V.J., Larsen, B., Bassett, D.S., Alexander-Bloch, A., Fair, D.A., Liston, C., Mackey, A.P., Milham, M.P., Pines, A., Roalf, D.R., et al., (2021). Neurodevelopment of the association cortices: Patterns, mechanisms, and implications for psychopathology. Neuron 109, 2820-2846.

Third, there are fairly large differences in the rCBF of subregions of the DMN, though one would expect them to be similar if they are maturing as a unit.

The DMN subregions identified with data driven approach from resting-state functional MRI (rs-fMRI) data and shown in current Figure 1—figure supplement 1 are consistent to the DMN subregions reproducibly identified by many other studies (see Raichle 2015 for review). The synchronized blood oxygenation level dependent (BOLD) signal from rs-fMRI BOLD has identified these subregions as a unit. Furthermore, despite various rCBF baseline values in these DMN subregions, the rCBF increase rates in these DMN subregions are consistent (Figure 3c). Despite that these DMN subregions are anatomically separate, it is reasonable to consider they are maturing functionally and physiologically together as a unit.

Raichle, M.E. (2015). The brain's default mode network. Annu. Rev. Neurosci. 38, 433-447.

Finally, the stronger correlation between increases in rCBF and rs-fcMRI metrics for the DMN may reflect its greater increase in rCBF during the time period studied. This may be due to the finding that VIS, located posteriorly and inferiorly, has already undergone its greatest increase in rCBF; while SM, located superiorly, has yet to undergo its greatest increase in rCBF. Thus both have relatively modest changes in rCBF during the period studied. The steeper rise in rCBF in the DMN, which may be by virtue of its location, would allow for a stronger correlation with maturational increases in rs-fcMRI parameters.

We thank the reviewer for this comment. Before original submission, we too were suspecting the stronger correlation between rCBF and rs-fcMRI metrics for the DMN only reflects greater increase in rCBF in the DMN. We added permutation tests with 10,000 permutations exactly to rule out that coupling in the DMN ROIs was driven by greater increase in rCBF in those ROIs. The results shown in Figure 5 clearly demonstrated strong coupling only localized at in the DMN regions. Figure 5—figure supplement 2 provided counter example that despite greater increase in rCBF in the DMN regions, rCBF increase in the DMN is not coupled to the rs-fcMRI metrics of the VIS or SM networks. We believe the combined Figure 5 and Figure 5—figure supplement 2 are robust to demonstrate strong coupling is specifically localized in the DMN regions, but not SM or VIS regions. Please also see related response to R1-2.

We agree with the reviewer that the VIS has already undergone its greatest increase in rCBF before the studied infant development period focused in this manuscript. Our previous study (Cao et al., 2017) on rs-fcMRI suggested SM located both inferiorly and superiorly also has already undergone its greatest functional connectivity increase during perinatal development, before the infant developmental stage focused in this manuscript.

Cao, M., He, Y., Dai, Z., Liao, X., Jeon, T., Ouyang, M., Chalak, L., Bi, Y., Rollins, N., Dong, Q., et al. (2017). Early development of functional network segregation revealed by connectomic analysis of the preterm human brain. Cereb. Cortex. 27, 1949-1963.

Since the gradient in increasing rCBF during development appears to be based on relative location, it is important to consider whether the superior-inferior gradient might be a technical artifact related to applying labeling pulses to the neck, which is getting longer, or changes in flow velocities with age.

We thank the reviewer for bringing up this good point. It is unlikely that the superior-inferior is a technical artifact. This comment has been included as the Essential Revisions comment #1 by the editor. Please see our response to that comment.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

The authors have partly responded to the reviewers' comments in my opinion. Some points still seem to me to be questionable in the manuscript, whereas some answers provided by the authors in their replies would have benefited from being added to the manuscript.

We agree with the reviewer’s comments. We did not put all responses into the manuscript since in eLife response letters are also published along with the manuscript. But in this revision, following this reviewer’s suggestion, these responses have now been added to the manuscript. Please see below for details.

– Essential revisions 1 and 2: unless I am mistaken, these elements have not been added to the manuscript. Perhaps this would be relevant for readers who don't read the reviewers' comments and the authors' answers in addition to the article.

We thank the reviewer for this comment. We have added our responses to previous essential revisions 1 and 2 to the Discussion section.

– R1-1: The answer provided might seem insufficient. The grey matter masks shown seem to me to clearly contain white matter, at least in the youngest infants for whom the cortex is very thin. The problem of reliably identifying the cortex to limit partial volume effects is definitely complex, and pCASL acquisition with thick slices doesn't avoid this problem anyway. That said, I still trust the measures performed, but I think it is important to discuss this point so that the reader is aware of this potential measurement bias. One way of doing this would be to estimate what the bias might be as a function of age (based on the estimated proportions of grey/white matter in the masks considered, and estimated rCBF in each tissue).

We agree with the reviewer that partial volume effects are a consideration factor for rCBF maps especially for smaller brain and thinner cortex of younger infants. Elaborated in our response to R1-1 in the last revision, we have adopted a method making the rCBF measure in the cortical mask in the template space consistent across ages for accurate and reliable voxel-wise comparison across age. As this reviewer suggested, the partial volume effect could be heterogenous to different ages with thinner cortex of younger infants. To address that, individual phase contrast (PC) MRI of all infants were obtained in our study to calibrate all rCBF maps of individual infants obtained with ASL scans by applying a normalization factor. In this way, the possible underestimation of rCBF in the younger infants due to more severe partial volume effects can be corrected. A function of age suggested by this reviewer to calibrate partial volume effects could also be effective but might require gold-standard infant gray matter rCBF measurements at different age as well as gold-standard segmentation of gray and white matter of infants. The former data is what we aim to obtain. The latter is known to be difficult and lacks consensus especially for younger infants due to poor T1-weighted or T2-weighted contrasts. Relevant discussion has been added in the last paragraph of the Discussion section. Please also see our response to RR2-3.

– R1-2: Again, the answer provided seems incomplete. When two measures depend on the same variable (age), their correlation can be partly driven by the common correlation, and it cannot then be considered as specific. The similarity between figures 3b and 5d on the lateral side draws attention to this point. Unless I am mistaken, the measures of functional connectivity in the visual and sensorimotor networks are much less age-dependent than in the DMN network (Figure 1-sup2), whereas the rCBF is age-dependent in all regions: this difference might drive the difference in FC-rCBF correlation observed across networks. A simple way to "prove" that the FC-rCBF correlation is indeed specific for the DMN would be to perform all these correlation analyses after removing all age effects in all measurements (region-wise or vertex-wise). These results would be convincing.

We fully understand the reviewer’s concern, as this was exactly our concern which motivated us to use the permutation tests to test whether the functional connectivity (FC)-rCBF correlation is due to their joint dependence on age. This permutation tests resulting in Figure 5 were elaborated in our response to R1-2 in last revision. Following the suggestion in this comment, we conducted further analysis testing the FC-rCBF correlation in the DMN regions after removing all age effects and found that, after regressing out the age effects, the correlation remains significant in both 0-12-month (r = 0.303, p < 0.001) and 12-24-month (r = 0.217, p < 0.001) groups. This result, combined with Figure 5, convincingly indicates that our finding of coupling between FC and rCBF cannot be explained just by age-related increases of both FC and rCBF in the DMN regions. The new result has been added in the subsection Coupling between rCBF and FC within DMN during infant brain development in the Results section.

– R1-4: it seems to me that the order of the method paragraphs should be changed to follow the new order of the results.

We thank the reviewer for this comment. The order of the paragraphs has been reorganized accordingly in the Materials and Methods section.

– R1-11: citation of the Cusack et al. paper would be welcome in the manuscript.

As suggested, citation of the Cusack et al. paper has been added to the subsection Data acquisition in Materials and Methods section.

– Introduction p3: the sentence "there have been no known whole-brain mappings of heterogeneous infant brain regional cerebral blood flow (rCBF) changes across landmark infant ages thus far" is incorrect (see the Lemaitre et al. paper pointed out in the previous review).

We thank the reviewer for this comment. We have revised the sentence in the Introduction section.

Reviewer #2 (Recommendations for the authors):

Thank you for the detailed answers to the questions and amendments to the manuscript. Whilst I agree with some of their responses, I feel there are still some unresolved issues which I think are fundamental to resolve:

We thank the reviewer for this comment. It seems some remaining issues are due to lack of clarification or miscommunication (e.g. same post-label delay (PLD) instead of misunderstood different PLD). As can be seen from our responses to these remaining comments, we explicitly acknowledged limitations of relatively lower pCASL resolution related to partial volume effects. All our imaging parameters were carefully selected and well justified. More importantly, for measurement accuracy all infant pCASL rCBF maps were calibrated by extra acquired subject-specific phase contrast MRI. Please see our responses to individual points below.

1. The use of similar acquisition parameters for all age groups: as far I understand it, the method proposed by Aslan et al. using PCA to calculate the labelling efficiency still requires the T1 of arterial blood as a parameter – for which neonatal values (according to their references) have been used. It would be good to know what would happen if more developmentally appropriate values were used in their CBF estimations. Secondly, the Aslan method provides a single corrective factor (and estimate of the labelling efficiency) for the CBF estimation based on the PCA measurement (which is global). Therefore it cannot correct for any regional effects of using a different post-label delay in the ASL acquisition – with knock-on effects for the regional CBF estimation (such as the spatial bias leading to the inferior-superior gradient for example). Would it be possible for the authors to provide some empirical evidence that the post-label delay has not influenced their results?

We thank the reviewer for this comment. We would like to point out that according to the rCBF calculation equation in the ASL white paper (Alsop et al., 2015), similar to the effect of labeling efficiency parameter on the rCBF map, the effect of arterial blood T1 on rCBF map estimation is global. By making averaged rCBF from pCASL equal to global CBF obtained from extra acquired phase contrast MRI of each infant using a "subject-specific" normalizing factor, we calibrated pCASL rCBF maps of all individual infants in this study. With this “subject-specific” calibration approach, the effects of individual variations in processing parameters such as T1 of arterial blood or labeling efficiency in the final estimation of rCBF measurement were largely minimized. However, for studies without extra phase contrast MRI acquired for "subject-specific" rCBF calibration, utilizing developmentally appropriate (as suggested by this reviewer) or even subject-specific T1 values of arterial blood is encouraged. We added discussion on the effect of T1 of arterial blood in rCBF estimation in the last paragraph of the Discussion section.

Secondly, according to the ASL white paper (Alsop et al. 2015), selected PLD should be larger than the arterial transit time (ATT) for reducing regional CBF bias. In this study by using a consistent PLD of 1600 ms larger than ATT of infants, the effect of PLD on rCBF differences across subjects or brain regions is relatively trivial. Please note the mean ATT of infants was reported to be less than 1500 ms (Varela et al. 2015; Kim et al. 2018). We also respectfully point out that we used in ASL acquisition a consistent PLD instead of a different PLD indicated in this comment. We acknowledge that due to lack of consensus on ASL acquisition parameters for the infant population, further studies focusing on systematically optimizing ASL acquisition protocol in the infant population are needed. Relevant details and discussion have been added to the subsection Data acquisition in Materials and Methods section.

Alsop, D.C., Detre, J.A., Golay, X., Günther, M., Hendrikse, J., Hernandez‐Garcia, L., Lu, H., MacIntosh, B.J., Parkes, L.M., Smits, M., et al. (2015). Recommended implementation of arterial spin‐labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med. 73, 102-116.

Kim, H.G., Lee, J.H., Choi, J.W., Han, M., Gho, S.M., and Moon, Y., (2018). Multidelay arterial spin-labeling MRI in neonates and infants: cerebral perfusion changes during brain maturation. American Journal of Neuroradiology 39, 1912-1918.

Varela, M., Petersen, E.T., Golay, X., and Hajnal, J.V., (2015). Cerebral blood flow measurements in infants using look–locker arterial spin labeling. Journal of Magnetic Resonance Imaging 41, 1591-1600.

2. With regards to the inferior-superior gradient, the references that the authors cite do not show the same kind of effects. Whilst many of them do suggest that CBF is higher posteriorly and is higher in the occipital lobe – it is not a "hot spot" in the same way as seen in the maps here (especially at the youngest ages). The maps in the Lamaitre work for example look very different (see figure 2 in their paper) to those shown in figure 3 here.

We agree with the reviewer that rCBF increase pattern is more complicated than relatively simplified inferior-to-superior or posterior-to-anterior gradient. Our Figure 3 suggests that rCBF increase pattern also presents a primary-to-association gradient reproducibly found in the literature (see Sydnor et al., 2021 for review). The differences between our Figure 3 with Figure 2 of Lemaitre Cereb Cortex 2021 might result from the two factors. First, the age range is different. In the same age range from 0 to 12 months, the rCBF maps are highly similar. The 12-month rCBF map in Lamaitre study also presents relatively higher rCBF in the occipital lobe, lateral temporal cortex, and lateral/medial prefrontal cortex. The “hot spot” appearance is more apparent in the 18-month and 24-month rCBF maps which are not available in Lemaitre study and are therefore not comparable. Second, although the pCASL perfusion MRI acquisition protocol we adopted in this manuscript is state-of-the-art, the voxel size is slightly larger than that of Lemaitre study. We believe slightly larger voxel size may contribute to blurring the rCBF map resulting in a “hot spot” in the inferior part of the brain. Using larger voxel size was by our design to ensure sufficient signal-to-noise-ratio (SNR) of the rCBF map. A 3T scanner used in this study rather than a 1.5T scanner in Lemaitre study also helps improve SNR for more accurate rCBF measurement, as seen in Figure 2 in this manuscript. Future studies with higher resolution and higher SNR are warranted to delineate the finer details of rCBF increase pattern in infants. The last paragraph in the Discussion section has been revised accordingly.

Sydnor, V.J., Larsen, B., Bassett, D.S., Alexander-Bloch, A., Fair, D.A., Liston, C., Mackey, A.P., Milham, M.P., Pines, A., Roalf, D.R., et al., (2021). Neurodevelopment of the association cortices: Patterns, mechanisms, and implications for psychopathology. Neuron 109, 2820-2846.

3. I am unconvinced by the argument that calculating grey matter maps in native high-resolution space and registering them together with the low-resolution CBF map into standard space thresholded probabilistic maps resolve the partial voluming issue. Fundamentally this does not change the base acquisition voxel which is very large and is the same across all of the infants studied – so differences relative to the size of the brain/cortex and partial voluming in the acquisition itself in the acquisition voxel are surely still an issue?

We thank the reviewer for this comment. We agree with the reviewer that partial volume effects of relatively lower resolution pCASL acquisition could be heterogeneous for infants at different ages as smaller brain and thinner cortex of younger infants cause larger partial volume effects when the imaging parameters are consistent across infants. Please note it is not the process of low-resolution rCBF map registered into standard space thresholded probabilistic maps that resolves the partial voluming issue. Instead, it is calibration with individual phase contrast (PC) MRI that contributes to correcting for bias of heterogenous partial volume effects. Please see details in our response to RR1-2. We also explicitly acknowledged heterogenous partial volume effects in the last paragraph of the Discussion section.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    MDAR checklist
    elife-78397-mdarchecklist1.docx (56KB, docx)

    Data Availability Statement

    Maps including regional cerebral blood flow (rCBF) maps and functional connectivity maps from multimodal infant MRI datasets including pseudo-continuous arterial-spin-labelled perfusion MRI and resting-state MRI of forty-eight infants are publicly available from Huang lab GitHub repository (https://github.com/haohuanglab/infant_perfusion_function, copy archived at swh:1:rev:34071e49232d7960ff6b78464b55077da136d868). However, we cannot openly share the raw, unprocessed MRI data, because The Institutional Review Board of Beijing Children's Hospital Research Ethics Committee (Approval number 2016-36) specifies the participants did not give consent for these data to be released publicly. The raw MRI data can be made available to individual researchers on informal request to the corresponding author through email hao.huang@pennmedicine.upenn.edu. Source code used in analysis and related documentation are also available at the Huang lab GitHub repository (https://github.com/haohuanglab/infant_perfusion_function).

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES