Innate connectivity patterns drive the development of the visual word form area

What determines the functional organization of cortex? One hypothesis is that innate connectivity patterns, either structural or functional connectivity, set up a scaffold upon which functional specialization can later take place. We tested this hypothesis by asking whether the visual word form area (VWFA), an experience-driven region, was already functionally connected to proto language networks in neonates scanned within one week of birth. Using the data from the Human Connectone Project (HCP) and the Developing Human Connectome Project (dHCP), we calculated intrinsic functional connectivity during resting-state functional magnetic resonance imaging (fMRI), and found that neonates showed similar functional connectivity patterns to adults. We observed that (1) language regions connected more strongly with the putative VWFA than other adjacent ventral visual regions that also show foveal bias, and (2) the VWFA connected more strongly with frontotemporal language regions than with regions adjacent to these language regions. These data suggest that the location of the VWFA is earmarked at birth due to its connectivity with the language network, providing evidence that innate connectivity instructs the later refinement of cortex.


Supplementary Information
Supplementary inventory: Supplementary Results 1&2: Using four fMRI tasks, subject-specific functional regions (fROIs) were identified in an independent group of adult subjects within the functional parcels used in the main analyses. fROIs across subjects demonstrate variability in spatial location across subjects and demonstrate the need for larger parcel regions that will certainly encompass the sites of functional specificity in the neonates in the main study. Percent signal change was also extracted in independent runs to demonstrate functional specificity of these fROIs.
Supplementary Results 3: Two-way mixed design ANOVA of language regions' FC using size as a covariate and two-way mixed design ANOVA of VWFA's FC using size as a covariate.  the MD localizer with Hard and Easy task conditions of a spatial working memory paradigm5; and the Dynamic localizer consisting of movie clips of faces, bodies, objects, and scenes6, 7. Data were collected on a 3T Siemens scanner with a 32-channel head-coil. Acquisition parameters were similar or identical to those used in the published studies for each localizer: the VWFA localizer was acquired with 2mm3 resolution, 2s TR, 30ms TE, 90° flip, 100x100 base resolution, 25 slices approximately parallel to the base of the temporal lobe to cover the entire inferior temporal cortex. We additionally collected a field map for distortion correction with the same slice prescription as the fMRI sequence (25 slices, 2mm3 resolution, 500ms TR, 55° flip, 100x100 base resolution). The Language & Speech and MD localizers were acquired with 2s TR, 30ms TE, 90° flip, GRAPPA acceleration factor 2, 2x2mm in-slice resolution, 4mm slice thickness, 31 slices for full-brain coverage; and the Dynamic localizer was acquired with 2TR, 30ms TE, 90° flip, 3mm3 voxels, and 32 slices for full-brain coverage. Experiments were counter-balanced across participants. A high-resolution (1 mm3) three-dimensional magnetization-prepared rapid acquisition with gradient echo (MPRAGE) scan was also acquired in all participants. All the imaging data were analyzed using standard pre-processing steps with Freesurfer (www.surfer.nmr.mgh.harvard.edu/) and FsFast (www.surfer.nmr.mgh.harvard.edu/FsFast/). Images were motion-corrected, de-trended, and fit using a standard gamma function (d = 2.25 and t = 1.25). Runs were registered to each subject's anatomical image using Freesurfer's bbregister.
We applied the functional parcels used in the main analyses to identify subject-specific functional regions (fROIs) using the same GSS approach and contrasts as outlined in the Methods. Individual fROIs were defined using the top 10% most active voxels within these parcels for each contrast of interest: words > line-drawings of objects for the VWFA, line-drawings of faces > objects for face regions, English > nonword sentences for language regions, nonword sentences > texturized speech for speech region, hard > easy for MD regions, movie clips of scenes > objects for scene regions and objects > scrambled objects for object regions. Each subject's fROI was registered to Freesurfer's CVS_MNI152 standard space and probabilistic heat maps were created based on these adults (shown in Supplementary Figure   1). The figure shows that these functionally selective fROIs are somewhat consistent across participants, where more than half of subjects' (>=7) fROIs land within similar spatial locations; however, we also observed large inter-subject variability as evidenced by the large spread of fROIs across e.g. temporal cortex for temporal language fROIs across subjects. These fROIs demonstrate variability in spatial location across subjects and demonstrate the need for larger parcel regions that will certainly encompass the sites of functional specificity in the neonates in the main study.

Supplementary Figure 1 | Probabilistic maps created based on subject-specific functional regions.
Heatmaps show voxels that show functional selective response to a given contrast in at least 1 subject (prob. = 0.07) to equal or greater than 6 subjects (prob. = 0.5). Contrasts of interest for each region are provided in the text and below each region; note that A1 for each individual was anatomically defined in Heschl's gyrus (superior and transverse temporal cortex from the FreeSurfer Desikan-Killiany parcellation). Engl, English sentences; Nonsn, non-word sentences; txtre, texturized; Src. Objects, scrambled objects. These fROIs demonstrate variability in spatial location across subjects and demonstrate the need for larger parcel regions that will certainly encompass the sites of functional specificity in the neonates in the main study.  Supplementary Results 3. We performed a two-way mixed design ANOVA of language regions' FC with age group (neonate, adult) as the between-group variable and target (VWFA, faces, scenes, objects) as the within-group variable, and the size of the functional parcels as a covariate. We found that even after accounting for size, the main effects of target (F(3,311) = 7.28, p < 0.001, partial η2 = 0.07), group (F(1,311) = 7.49, p = 0.007, partial η2 = 0.02), and the interaction (F(3, 311) = 5.65, p = 0.001, partial η2 = 0.05) were still significant. We also performed a two-way mixed design ANOVA with age group (neonate, adult) as the between-group variable and target (language regions, multiple-demand regions, speech, A1) as the within-group variable, and the size of the functional parcels as a covariate. Our results showed no main effect of age group (F(1,311)  Supplementary Figure 3 | Averaged voxel-wise functional connectivity maps within the ventral temporal cortex (VTC) using language regions as the seed. Functional connectivity maps were calculated (fisher's z transformed, z(r)) for each individual and then averaged across individuals. (a) all adults' maps were registered to the Freesurfer CVS_avg35_inMNI152 brain, and the averaged map across adults was projected to the surface (left) and volume (right) in the template space. (b) all neonates' maps were registered to 40-week (gestational age) template brain8, and the averaged map across neonates was overlaid on the neonate template volume (right). Note that we also projected neonates map to the adult template surface (left) with the caveat that potential distortions may at the voxel-wise level.

Supplementary Results 4.
We broke up the language regions into frontal and temporal components and performed the same FC analyses as in the main paper but comparing VWFA connectivity to the target regions grouped by temporal and frontal. We found that the VWFA was more connected with temporal language regions than speech and A1 regions in both neonates

Supplementary Figure 4 | Functional connectivity between VWFA (seed) and temporal and frontal regions. (a)
Mean FC between VWFA and regions in temporal (i.e., temporal language regions, speech, and A1). (b) Mean FC between VWFA and regions in frontal cortex (i.e., frontal language regions, multipledemand regions). Connectivity values were Fisher z transformed. Error bars denote s.e.m. Individual data points (n = 40 for each age group) were shown for each category. Horizontal bars reflect significant post hoc paired t-tests p < 0.05, corrected.
We also performed the same parametric voxelwise analyses as done for the VTC in the main paper, but here we used the VWFA as the seed and characterized its FC to temporal and frontal cortex. Consistent with parcel-based analyses, when parametrically increasing threshold of FCs, we found that voxels revealed higher FC to VWFA located in temporal and frontal cortex that happen to be in language regions

Supplementary Results 5.
We obtained association test maps from a meta-analysis approach (Neurosynth, https://neurosynth.org) for all functions of interest (terms used: language (1101 studies), speech (642), primary auditory (114), faces (864), objects (692), place (189), visual word (117)) with a threshold of z > 3, and took the intersection between the functional parcels used in the main analyses with these Neurosynth-generated maps. These maps shared similar spatial locations with our functional parcels, but yield more conservative regions for most of functional categories (Supplementary Figure 7a and 7b, also see Supplementary Table 1  Supplementary Results 6. We used ANTs to register parcels to neonates and adults because it's commonly used in developmental studies 9-11. We manually checked the registration results to ensure that functional parcels were in the relatively right locations (see Supplementary Figure 8a for example registration output for a representative adult and neonate). Next, for a quantitative analysis, we compared our registration accuracy to published registration methods used in a recent and comparable neuroimaging study of infants, where functional parcels were again defined based on fMRI data in a group of adults (and in fact using the same atlas/parcels that we are using here), and then registered to the infant brain but using FLIRT (FMRIB's Linear Image Registration Tool from FSL) instead of ANTs12, 13. We took the binary gray matter mask from CVS average-35 MNI152 template brain and then registered it to individual's native space with both ANTs and FLIRT. The registration result was compared to each individual's own binary gray matter mask. We found that in general, both methods yield high accuracy (over 90%), but ANTs significantly Connectivity values were mean-centered and averaged within each of the four categories to plot the relative patterns for the adult (n = 40) and neonate groups (n = 40). (d) Mean FC between VWFA-p and regions in temporal cortex (i.e., temporal language regions, speech and A1)). (e) Mean FC between VWFA-p and regions in frontal cortex (i.e., frontal language regions, multiple-demand regions). Connectivity values were Fisher z transformed. Error bars denote s.e.m. Individual data points (n = 40 for each age group) were shown for each category. Horizontal bars reflect significant post hoc paired t-tests p < 0.05, corrected.