Sequence Alterations of Cortical Genes Linked to Individual Connectivity of the Human Brain

Abstract

Individual differences in humans are driven by unique brain structural and functional profiles, presumably mediated in part through differential cortical gene expression. However, the relationships between cortical gene expression profiles and individual differences in large-scale neural network organization remain poorly understood. In this study, we aimed to investigate whether the magnitude of sequence alterations in regional cortical genes mapped onto brain areas with high degree of functional connectivity variability across individuals. First, human genetic expression data from the Allen Brain Atlas was used to identify protein-coding genes associated with cortical areas, which delineated the regional genetic signature of specific cortical areas based on sequence alteration profiles. Thereafter, we identified brain regions that manifested high degrees of individual variability by using test-retest functional connectivity magnetic resonance imaging and graph-theory analyses in healthy subjects. We found that rates of genetic sequence alterations shared a distinct spatial topography with cortical regions exhibiting individualized (highly-variable) connectivity profiles. Interestingly, gene expression profiles of brain regions with highly individualized connectivity patterns and elevated number of sequence alterations are devoted to neuropeptide-signaling-pathways and chemical-synaptic-transmission. Our findings support that genetic sequence alterations may underlie important aspects of brain connectome individualities in humans.

Significance Statement: The neurobiological underpinnings of our individuality as humans are still an unsolved question. Although the notion that genetic variation drives an individual’s brain organization has been previously postulated, specific links between neural connectivity and gene expression profiles have remained elusive. In this study, we identified the magnitude of population-based sequence alterations in discrete cortical regions and compared them to the brain topological distribution of functional connectivity variability across an independent human sample. We discovered that brain regions with high degree of connectional individuality are defined by increased rates of genetic sequence alterations; these findings specifically implicated genes involved in neuropeptide-signaling pathways and chemical-synaptic transmission. These observations support that genetic sequence alterations may underlie important aspects of the emergence of the brain individuality across humans.

Introduction

Individual differences in social-emotional-behavioral performance is one the most remarkable features of humanity. Although there are shared traits across individuals, other characteristics may represent distinct and unique personal qualities. However, the mechanisms through which the human brain promotes uniqueness across multiple domains remains ill-defined. It has been postulated that we think and behave differently due in part to individual differences in (1) the genetic variants or mutations in individual genomes and (2) the idiosyncratic conformations of individual human brain anatomy and connectivity (Mayr et al. 2005; Dickie et al. 2014). For example, reports have identified genetic variants as predisposing risk factors for developing neurodegenerative diseases (Albert and Kruglyak 2015; Zhang and Lupski 2015), yet little is known about the relationships between genetic variability and the normal range of broadly-distributed neural network connectivity (Dickie et al. 2014).

The most commonly used scientific approaches primarily aim to remove the influence of individualized neural circuit dynamics by using population or group-average analyses. Nonetheless, several neuroimaging studies have shown that large-scale functional and structural networks display individualized anatomy and connectivity (Mueller et al. 2013; Braga and Buckner 2017; Gordon et al. 2017; Peña-Gómez et al. 2017; Amico and Goñi 2018). Although the integration of subject-level neural system and genetic expression information remains challenging, the emerging field of neuroimaging-genetics allows for new opportunities to characterize the neurobiological basis of brain organization (Richiardi et al. 2015; Wang et al. 2015; Krienen et al. 2016; Bassett and Sporns 2017; Ortiz-Terán et al. 2017). In this study, we aimed to characterize the genetic underpinnings mediating variability in neural connectivity, which is an important feature of individuality (Dickie et al. 2014). We investigated individuality by integrating two important factors, the genotypic variance of cortical gene sequence alterations and the profiles of individual functional connectomes. First, we used human genetic expression data from the Allen Brain Atlas (Hawrylycz et al. 2012) to identify protein-coding gene expression profiles associated with cortical tissues, and subsequently delineated the regional genetic signature that most distinguished a given cortical area based on sequence alterations (Aken et al. 2016). Second, we identified brain regions that manifested high rates of individual variability by using test-retest functional connectivity MRI and graph-theory analyses. Using this latter approach, we identified discrete brain areas displaying stable subject-specific connectivity. Finally, we sought to combine cortical gene expression information with the topographic distribution of brain areas exhibiting individualized connectional variability. We hypothesized that the cortical areas with high levels of allelic sequence alterations would map onto cortical regions exhibiting elevated rates of individualized functional connectivity.

Methods

Assessment of Sequence Alterations in Cortical Genes

To obtain cortical gene expression profiles, we used the whole-brain genome-wide expression data from the Allen Human Brain Atlas (Fig. 1A). Specifically, we used the cortical surface representations of the transcription expression patterns for the 20 737 protein-coding genes analyzed in six human brains (mean age of 42.5, ranging from 24 to 57; 5 M/1 F; see Supplementary Table S1 for more detailed demographics). This data was downloaded from French and Paus 2015; who in turn used the complete microarray gene expression datasets from the Allen Institute of Brain Science website (http://human.brain-map.org/static/download/ downloaded in May 2015). The atlas is based on 5692 measurements of gene expression profiles in 3702 post-mortem brain samples (French and Paus 2015). The anatomical transformation is based on the 68 cortical regions of the Desikan–Killian atlas (Desikan et al. 2006), which covers the entire human cortex. Individual vectors of the median expression of protein-coding genes across brain regions were used to generate the final atlas onto which our analyses were mapped (Fig. 1A; letter a).

Figure 1.

Pipeline overview. (A) Diagram of genetic expression matrix for ~20 000 protein-coding genes from the Allen Brain Atlas covering the entire human cerebral cortex, based on the 68 regions of the Desikan–Killian atlas. This genetic expression data allows the visualization and analysis of single protein-coding gene expression levels across the cortex [a; green colors in the brain illustration), as well as the investigation of the whole transcriptome to identify which genes are highly expressed in discrete cortical regions (b; blue plot of z-scores and histogram red bars (>2 standard deviations)]. This strategy served to calculate specific genetic variant scores of the genetic profiles of brain regions. (B) Diagram of connectivity analysis in two theoretical individuals with two separate time points, in which, first, stable connections across test and re-test resting-state scans (represented as light blue links; spurious connections represented as black dashed links) were detected for each subject (subjects 1 and 2; black links). Then, stable common connectivity across the sample (represented as violet links) were disregarded to focus on the stable connectivity profiles across test and re-test scans that were distinct across individual subjects (represented as red links). Finally, a consensus map of regions with different degrees of individual links was obtained (right cortical map). (C) The spatial cortical relationship between genetic variant quantifications (represented in y-axis) and a brain connectivity phenotype containing the magnitude of connectivity individualities (or connectivity phenotype represented in x-axis) were computed using a linear analytical approach.

We first created a “cortical gene signature” for each Desikan–Killian cortical area by building histograms of their entire transcriptome and selecting genes that displayed the highest transcription expression levels, specifically identifying genes two standard deviations (2 SD) above the mean of all genetic data in a given region (Fig. 1A; letter b). Thus, in each region we only selected the 2.5% of the transcriptome information representing the greatest uniqueness per region. Once we obtained the genetic expression profiles that characterized cortical areas, we quantified the genetic sequence alterations in all the genes from the two-standard deviation results, using Ensembl (Eilbeck et al. 2005) (thus, genetic sequence alterations (also refer in the text as genetic variants) were identified based on large aggregate of individuals; Supplementary Table S2; Aken et al. 2016). Given the limited post-mortem samples, when combining high-resolution neuroanatomy and available gene expression data it is important to consider the consistency of gene expression patterns across both cerebral hemispheres. Accordingly, we obtained both the symmetric and asymmetric counts of highly expressed genes by taking into consideration the corresponding left and right cortical regions in each hemisphere of the Allen Human Brain Atlas aggregate data. As the symmetric genes were less numerous than the asymmetric ones, we normalized the symmetric and asymmetric genetic variant scores independently (scaling between 0 and 1) to equilibrate the weight of the symmetric profile, and subsequently averaged them. Finally, we used Caret software (PALS surface (PALS-B12); interpolated algorithm and multi-fiducial mapping) as the final cortical space visualization tool.

For quantifying the genetic variant scores associated to each protein-coding gene, we used the Ensembl website (Eilbeck et al. 2005). For the primary measure of allelic variance, we employed the annotation system of Sequence Ontology (SO) and its parent term sequence alterations to characterize the sequence alteration score in all our targeted protein-coding genes obtained from the previous step (same matching gene names as those provided by the Allen Human Brain Atlas genetic data). Secondary analyses examined the individual effects of six main components (children SO terms) of the sequence alteration parent, including: (1) single nucleotide variants (SNVs), (2) insertion of one or several nucleotides added between two adjacent nucleotides in the sequence (insertions), (3) deletion of one or several nucleotides (deletions), (4) an insertion and a deletion, affecting at least two nucleotides (indels), (5) a sequence variation where the length of the change in the variant is the same as that of the reference (substitutions), and (6) tandem repeat affecting two or more adjacent copies of region (tandem repeats) (Supplementary Fig. S1). Thus, we obtained seven different genetic variant scores (one parent and six individual components) for all representative genes of each brain region. This approach led to a single cortical map per each averaged genetic variant type that was logarithmically transformed to normalize intensity.

Assessment of Individual Level Functional Connectivity

We investigated the individual functional connectivity of healthy subjects (N = 40) recruited and enrolled with test-retest scanning administrated within approximately one month apart (mean age of 23.6, ranging from 18 to 48; 25 M/15 F; see Supplementary Table S3 for more detailed demographics). Apart from the main sample of the study, we also included an additional sample of individuals (N = 40; mean age of 25.2, ranging from 18 to 50; 21 M/19 F) for replication purposes of the network neuroimaging analysis (Supplementary Fig. S2). All subjects were without neurologic and psychiatric conditions, and had not previously taken psychoactive medications. The Brain Genomic Superstruct Project, a multi-laboratory neuroimaging collaborative effort at Harvard University, the Massachusetts General Hospital and the greater Boston area, provided this subject data (Holmes et al. 2015). Subjects provided written informed consent in accordance with Helsinki Declaration and guidelines set by institutional review boards of Harvard University and Partners Healthcare.

Scanning was performed on a 3-Tesla TimTrio system (Siemens) with a 12-channel phased-array head coil. High-resolution 3D T1-weighted magnetization multi-echo images for structural anatomic reference (multi-echo MPRAGE; TR, 2200 ms; TI, 1100 ms; TE, 1.54 ms for image 1–7.01 ms for image 4; flip angle, 7°; 1.2 × 1.2 × 1.2 mm; FOV, 230) and a gradient-echo echo-planar pulse sequence (EPI) sensitive to blood oxygenation level-dependent (BOLD) contrast for functional imaging data were obtained. EPI parameters were as follows: TR, 3000 ms; TE, 30 ms; flip angle, 85°; 3 × 3 × 3 mm voxels; FOV, 216; 47 slices (each run lasted 6.12 min, with 124 time points). During the scans, participants were instructed to remain still without falling asleep, and keep their eyes open while blinking normally.

We applied Data Processing Assistant for Resting State (DPABI) (Chao-Gan and Yu-Feng 2010) to the rs-fMR functional imaging data in accordance with Statistical Parametric Mapping functions (SPM12; http://fil.ion.ucl.ac.uk/spm), run in Matlab (v8.0, The Mathworks Inc., Natick, MA). Resting-state data preprocessing included: (1) removing the first four volumes of each run for T1-equilibration; (2) performing slice-timing correction; (3) using a rigid-body six-parameter linear transformation to re-align the functional volumes for each subject in order to correct for head movement; (4) co-registering individual structural images (T1-weighted images) to the individual mean functional image; (5) segmenting T1-weighted images into gray matter, white matter and cerebrospinal fluid (CSF); (6) removing sources of spurious variance from the functional data through linear regression including (a) 24 motion-related parameters derived from volume-realignment (Friston et al. 1996; Satterthwaite et al. 2012), (b) signal averaged only from cerebral white matter and lateral ventricles CSF and (c) polynomial quadratic trend; (7) applying spatial transformation to MNI space using individual segmentations of each participant’s high-resolution anatomical image; (8) conducting spatial smoothing (isotropic Gaussian kernel of 6-mm FWHM); (9) applying temporal band-pass filtering (0.01–0.08 Hz) to diminish the effect of low-frequency drift and high-frequency noise (Biswal et al. 1995; Lowe et al. 1998) and finally (10) using scrubbing of image volumes with excess head motion (frame displacement >0.5 mm) through spline interpolation (Jenkinson et al. 2002).

To identify the brain areas with high degree of individualized functional connectivity, we performed a whole-brain graph-theory analysis (see diagram in Fig. 1B illustrating a N = 2 case example for a single star network). First, we computed the Pearson’s r correlation coefficient between pairs of voxels across the whole brain using the time course of low-frequency BOLD fluctuations from test and retest scans (represented as Time 1 and Time 2 in Fig. 1B). A whole brain mask of 4652 voxels at 8 mm isotropic voxel size (n) was used to obtain a final n × n association matrix for each individual (Ortiz-Terán et al. 2017). Second, we used both the whole spectrum of correlation values without introducing any cutoff (including negative and positive values) to calculate the temporal stability of functional connections between the test-retest scans (represented as light blue connectivity in Fig. 1B). We estimated a matrix of stability scores (T) in which for each functional connection of each individual we obtained the mean correlation value between the two-time points divided by their absolute difference (Equation 1). Third, we identified if each stable functional connection in each subject was specific for that individual (represented as red connectivity in Fig. 1B) or alternatively pertained to a network shared across the sample (common connectivity; represented as violet connectivity in Fig. 1B). We obtained the mean (μ) and standard deviation (σ) of the sample for each possible connection in the brain mask. Then, we converted the values of each subject-level stability scores matrix (T) to z-scores using the previously obtained sample-based mean and standard deviation (Equation 2). We used the normal cumulative distribution to obtain right-tailed P-values associated to each z-score to denote the connections that occurred differentially in each individual compared to other subjects (Equation 3). Then, P-values were corrected using a false discovery rate threshold (FDR) at a q level of 0.01 (Equation 4) (Benjamini and Hochberg 1995). All connections with P-values above the corrected statistically significant threshold were considered stable individual connections, while all connections with P-values below that level were considered as part of a common network shared by the sample, and therefore were disregarded for the purposes of this study. Finally, we used a graph-theory strategy to sum all temporally stable individual connections in each node to obtain a degree centrality connectivity map (Equation 5). Thus, this approach generated cortical maps in which cortical regions that accumulated different levels of stable connections can be detected (right brain illustration in Fig. 1B), in which regions with high degree centrality of stable connectivity can be interpreted as hub regions for individualized stable connectivity. We averaged all degree centrality maps of subjects (scaling between 0 and 1) and converted from voxel-level to 68 Desikan–Killian regions for comparison purposes with the cortical estimation of gene expression variability.

$$\mathrm{T}=\frac{(M_1+M_2)/2}{abs(M_1-M_2)}$$ (1)

where M₁ and M₂ are the test and retest correlation matrices, respectively.

$$Z_{ijk}=\frac{T_{ijk}-\mu }{\sigma }$$ (2)

where k is the index of each subject, and μ and σ are the mean and standard deviation of each connectivity value in T across the sample.

$$P_k=1-F\left(Z_k|0,1\right)=1-\frac{1}{\sqrt{2\pi }}{\int }_{-\infty }^{Z_k}{e}^{\frac{-t^2}{2}}\mathrm{d}t$$ (3)

$$P_{\mathrm{t}\mathrm{h}\mathrm{r}\_\mathrm{f}\mathrm{d}\mathrm{r}}=\mathrm{m}\mathrm{a}\mathrm{x}\left(P\le \frac{m*n}{I^T*q*{\sum }_{i=1}\frac{1}{I_i}}\right)$$ (4)

where $I=(\mathrm{1,2,3}\cdots m*n)$ and q = 0.01

$$\mathrm{W}{\mathrm{D}}_i=\sum _{j=1}{Z}_{ij}[p_{ij}<p_{t\mathrm{h}\mathrm{r}\_\mathrm{f}\mathrm{d}\mathrm{r}}]$$ (5)

where Z is the stable connectivity matrix corrected by FDR.

Spatial Intersection of Cortical Gene Sequence Alterations and Brain Connectivity Individualities

For the primary analysis, we computed the cortical spatial similarity between the magnitude of a given genetic sequence alteration and the individualized functional connectivity map for Desikan–Killian areas using a linear analytical approach (Fig. 1C). Note, this linear correlation approach is used as a measure of similarity in which the P-value is not consider, as brain regions are not spatially independent. To calculate the significance of our similarity maps between individualized functional connectivity and the seven genetic variants maps we built null hypothesis distributions using a permutation strategy (random kernel; 1000 iterations) that randomly resampled the data of cortical vectors and created a reference histogram. We performed the same procedure for all similarity assessments. To control for multiple testing, we used a FWE P < 0.01 correction to correct all P-values obtained from the seven similarity analyses of genetic variants vs. brain connectivity maps.

Finally, we used a two-dimensional K-means clustering approach to characterize the neurobiological processes associated with different levels of sequence alteration and individualized functional connectivity map associations. We first applied a Silhouette method to estimate the appropriate number of clusters of the data. This allowed us to find five clusters of brain regions in which we could combine their corresponding genes and investigate significant Gene Ontology (GO) functional profiles (Ashburner et al. 2000; Gene Ontology Consortium 2015). Thus, we used GO over-representation technique based on predefined gene classifications to assign functional characteristics to our genetic profiles of brain clusters. As a final step, we obtained the GO profiles (FWE P < 0.01 correction) for the common genetic background of all clusters (center/pentagon set in Venn diagram) and for each cluster profile without the common background (specific sets of clusters and overlapping sets other than the common background in Venn diagram).

Results

We found that the individualized functional connectivity and the sequence alterations in cortical genes displayed a strong spatial similarity (Fig. 2A; main scatterplot on top-left; similarity coefficient = 0.377; R² = 14.2%.). Sequence alterations scores were statistically significantly related to the topographic distribution of connectivity individualities in the human cortex (Fig. 2A; P = 0.0011; significant at FWE P < 0.01 correction; Fig. 2B). By contrast, discrete allelic variants, such as SNVs, deletions or insertions, only displayed a trend in their spatial relationship with the individualized functional connectivity map (Supplementary Fig. S3).

Figure 2.

(A) Scatterplot showing the spatial similarity relationship throughout brain regions between the sequence alterations and individualized functional connectivity maps (linear fit represented with a dotted line). Color code represents a two-dimensional k-means clustering partition of brain regions in five clusters derived from a Silhouette analysis (line plot in inset). (B) Null hypothesis distribution of the similarity between sequence alterations and individualized functional connectivity maps based on a resampling random permutation strategy. One, two and three standard deviation (σ) thresholds are highlighted with a red dotted line. Black arrow marks the similarity coefficient and significant P-value of the sequence alterations and individualized functional connectivity map association. (C) Left and right cortical hemispheric surface of genetic sequence alterations and individualized functional connectivity maps (lateral, medial and flat projections) in jet color scale (min=2% and max = 98% threshold visualization), in which a and b point to the occipital and temporal lobe, respectively. (D) Left and right cortical hemispheric surface of cluster 1–5 from k-means clustering partition in (A) (lateral, medial and inflated projections), in which c and d point to the visual and auditory primary cortex (dark and light blue) and e points to heteromodal cortex (red and orange).

In parallel, we identified the genetic sequence alterations scores for each brain region (Fig. 2C; cortical maps on left column). Visual cortex, particularly primary areas, and dorsomedial parietal and frontal regions displayed low sequence alterations (Fig. 2C; letter a in flat in left column cortical surface); other primary systems such as somatosensory and auditory networks showed an intermediate level of degree of genetic sequence alterations. Heteromodal cortices in temporo-inferior areas, insula, posterior cingulate and, to lesser extent, in lateral fronto-parietal regions expressed high levels of sequence alterations (Fig. 2C; letter b in flat in left column cortical surface). We also characterized the brain regions with high degree of individualized functional connectivity (Fig. 2C; cortical maps on right column). Visual cortex and primary auditory areas displayed low individualized functional connectivity (Fig. 2C; letter a and star symbol in flat in right column cortical surface), somatosensory regions showed an intermediate level of individualized connectivity, and heteromodal cortices in lateral fronto-parietal regions and inferior temporal areas displayed highly individualized connectivity centrality (Fig. 2C; letter b in flat in right column cortical surface).

We also observed that brain regions clustered into five sets based on their association between the sequence alterations and individualized functional connectivity maps (Fig. 2A and D; color code in the scatterplot comes from the optimal partition outcome of the Silhouette analysis, line graph). Although isolated sequence alterations and individualized functional connectivity distributions showed trends towards hierarchically ordered cortical systems when treat them separately (Fig. 2C), a robust spatial gradient from low—in primary visual and auditory cortices (Fig. 2D; dark-light blue in all cortical surfaces)-, moderate levels—in somatomotor and dorsomedial fronto-parietal (attentional) cortices (Fig. 2D; green in all cortical surfaces)-, to high levels of individualities—in fronto-parietal and temporal heteromodal cortices, including medial and temporal pole areas (Fig. 2D; red–orange in all cortical surfaces)—became evident when using the composite map of clusters (Fig. 2D). Importantly, we found that these five clusters of brain regions displayed genetic profiles that are devoted to specific Gene Ontology (GO) functions (Fig. 3; gene lists attached in Supplementary Material). While the common genetic background to all five clusters is linked to multicellular organismal processes, a non-neural specific function (Fig. 3A; center of the Venn diagram), genes from Cluster 1 to 5 displayed neuro-specific functionality. For instance, genetic profiles of Cluster 1 and 2 (Fig. 3A; dark-light blue) were concordantly over-represented by sensory perception processes (FWE P < 0.01; <2-fold) and brain regions with both high degree of individualized functional connectivity and genetic sequence alterations collected in Cluster 4 and 5 (Fig. 3A; orange–red) showed a genetic profile with over-representations in neuropeptide-signaling pathway (FWE P < 0.01; >2-fold) and chemical-synaptic transmission (FWE P < 0.01; <2-fold). Among non-neural specific functions, significant over-representation was observed in cyclic nucleotide, G-protein and ERK1/ERK2 cascade processes (FWE P < 0.01; >2-fold) particularly in Cluster 5, among other examples.

Figure 3.

(A) Cortical surface with clustering partition from Fig. 2D and Venn diagram displaying the common (dotted line) and overlapping genetic profiles of the five clusters that characterized the sequence alteration and individualized functional connectivity map relationships. (B) Fold over-representation of Gene Ontology biological process profiles of the five clusters that characterized the sequence alteration – individualized functional connectivity map relationships (all annotations are significant at the corrected level of FWE P < 0.01). Neuro-related annotations were highlighted with the same color code as clusters in Fig. 3A. Two-fold over-representation threshold was pointed with a red dotted line.

Discussion

Individual difference in humans, such as cognition, social-affective functions, and behavior, arise from different neural circuit configurations and neuroplasticity mediated by every-day experiences (Mayr et al. 2005; Freund et al. 2013). As such, each unique brain might hypothetically show features that deviate from collective species-wise neural organizational principles. Currently, however, how the brain fosters individualized connectivity profiles remains unknown. In this combined neuroimaging-genetics study, we report evidence supporting that genetic sequence alterations are significant factors facilitating the emergence of individualized brain connectomes. That is, a wide diversity of genetic mutations in cortical genes is needed for the assembly of our individualized neural networks.

In this study, we identified a convergence between cortical regions expressing protein-coding genes with a high degree of genetic variability as measured by sequence alterations and brain areas exhibiting a high degree of individualized functional connectivity differences. A brain gradient increasingly co-localized the degrees of genetic sequence alterations and individualized functional connectivity maps from primary modal to heteromodal/association cortices in lateral fronto-parietal, and anterior, medial and inferior aspects of the temporal lobe. Thus, central aspects of our individual connectome organization appear related to a high concentration of genetic variants allocated in regions for higher-order cognitive processing and reasoning abilities as previously linked to individual human qualities (Olson et al. 2007; Hariri 2009; Aichelburg et al. 2016). Moreover, additional evidence about this specialization gradient comes from the specific genetic profiles devoted to distinctive GO neurobiological processes that the spatial cortical interplay between genetic variability and connectivity individuality highlights. For instance, low degrees of both sequence alterations and individualized functional connectivity were observed in primary visual and auditory cortices, which in turn show genetic profiles with GO functionality devoted to sensory processes. In comparison, high degrees of both sequence alterations and individualized functional connectivity in heteromodal cortices were associated with the significant expression of neuropeptide-signaling pathway and chemical-synaptic transmission genes. Notably, both phenomena have been described as central biological processes involved in the individuation of the nervous system (Tian et al. 2014). Neuropeptides are varied signaling molecules generated after the cell surface receptor binding of a peptide neurotransmitter, and have been linked to the regulation and flexibility of individual behavior by modulating neural circuits (in C. elegans; Jarrell et al. 2012; Izquierdo and Beer 2013; Leinwand and Chalasani 2013). Neuropeptide signally pathways are also involved in aspects of membrane excitability, neural communication, synaptic transmission, synaptic growth development (in Drosophila), and learning and memory processes (Chen and Ganetzky 2012; Leinwand and Chalasani 2013). In addition, we found three none GO neuro-related biological processes—cyclic nucleotide, G-protein and ERK1/ERK2 cascade profiles. However, those cellular pathways are also commonly recognized as mediators in the regulation of cell growth and adhesion, as well as neuronal and neuropeptide-signaling.

There are several limitations in this study. The dataset provided by the Allen Brain Atlas only includes six subjects, however, this is the only publicly available dataset with human neural genetic expression co-registered with high-resolution MRI. Although the ideal dataset would consist on several brain MRIs performed during the lifespan of a subject, along with post-mortem mapping of cortical gene expression profiles, this dataset does not exist today. Additional factor requiring future considerations include epigenetic modifications and gene x environment influences more broadly. An important strength of this study is the replication of our findings using independent neuroimaging datasets.

In conclusion, we report evidence for a gradient of genetic-functional connectivity relationships in the cortical mantle supported by integrating data from gene expression profiling, allelic variations, connectomics, and gene set enrichment analysis for biological interpretation. The findings suggest that sequence alterations in specific brain regions of the human brain (particularly in the heteromodal cortex) play an important role in shaping individualized functional connectivity profiles. This phenomenon points to the neurobiological basis for the emergence of individual differences in the human brain, and consequently elucidates aspects of the individualized characteristics of human beings.

Funding

This work has been partially supported by the National Institutes of Health [grant K23EB019023 (J.S.) and grant 2T32EB013180-06 (L.O.T.) from the National Institute of Biomedical Imaging and Bioengineering (N.I.B.I.B.); and grant K23MH111983 (D.P.) from National Institute of Mental Health (N.I.M.H.)], the Basque Country Government [Post-Doctoral Fellowship Program (I.D.)] and the Sidney Baer jr. Foundation (D.P.).