Comprehensive transcriptional map of primate brain development
Trygve E. Bakken,1,*Jeremy A. Miller,1,*Song-Lin Ding,1,*Susan M. Sunkin,1Kimberly A. Smith,1Lydia Ng,1Aaron Szafer,1Rachel A. Dalley,1Joshua J. Royall,1Tracy Lemon,1Sheila Shapouri,1Kaylynn Aiona,1James Arnold,1Jeffrey L. Bennett,2Darren Bertagnolli,1Kristopher Bickley,1Andrew Boe,1Krissy Brouner,1Stephanie Butler,1Emi Byrnes,1Shiella Caldejon,1Anita Carey,1Shelby Cate,1Mike Chapin,1Jefferey Chen,1Nick Dee,1Tsega Desta,1Tim A. Dolbeare,1Nadia Dotson,1Amanda Ebbert,1Erich Fulfs,1Garrett Gee,1Terri L. Gilbert,1Jeff Goldy,1Lindsey Gourley,1Ben Gregor,1Guangyu Gu,1Jon Hall,1Zeb Haradon,1David R. Haynor,3Nika Hejazinia,1Anna Hoerder-Suabedissen,4Robert Howard,1Jay Jochim,1Marty Kinnunen,1Ali Kriedberg,1Chihchau L. Kuan,1Christopher Lau,1Chang-Kyu Lee,1Felix Lee,1Lon Luong,1Naveed Mastan,1Ryan May,1Jose Melchor,1Nerick Mosqueda,1Erika Mott,1Kiet Ngo,1Julie Nyhus,1Aaron Oldre,1Eric Olson,1Jody Parente,1Patrick D. Parker,1Sheana Parry,1Julie Pendergraft,1Lydia Potekhina,1Melissa Reding,1Zackery L. Riley,1Tyson Roberts,1Brandon Rogers,1Kate Roll,1David Rosen,1David Sandman,1Melaine Sarreal,1Nadiya Shapovalova,1Shu Shi,1Nathan Sjoquist,1Andy J. Sodt,1Robbie Townsend,1Lissette Velasquez,1Udi Wagley,1Wayne B. Wakeman,1Cassandra White,1Crissa Bennett,1Jennifer Wu,1Rob Young,1Brian L. Youngstrom,1Paul Wohnoutka,1Richard A. Gibbs,5Jeffrey Rogers,5John G. Hohmann,1Michael J. Hawrylycz,1Robert F. Hevner,6Zoltán Molnár,4John W. Phillips,1Chinh Dang,1Allan R. Jones,1David G. Amaral,2Amy Bernard,1 and Ed S. Lein1
1Allen Institute for Brain Science, Seattle, Washington 98109, USA
2Department of Psychiatry and Behavioral Science, California National Primate Research Center, The M.I.N.D. Institute, University of California, Davis, Sacramento, CA 95817, USA
3Department of Radiology, University of Washington, Seattle, Washington 98195, USA
4Department of Physiology, Anatomy and Genetics, University of Oxford, South Parks Road Oxford OX1 3QX, UK
5Human Genome Sequencing Center and Department of Molecular and Human Genetics, Baylor College of Medicine, Houston, Texas 77030, USA
6Center for Integrative Brain Research, Seattle Children’s Research Institute, Seattle, Washington 98101, USA
Corresponding author: Ed S. Lein, Ph.D., Allen Institute for Brain Science, 615 Westlake Ave N, Seattle, WA 98109, Tel: 206.548.7039, Fax: 206.548.7071, Email: gro.etutitsninella@LdE
*These authors contributed equally
Author information ►Copyright and License information ►
The publisher's final edited version of this article is available at Nature
See other articles in PMC that cite the published article.
The transcriptional underpinnings of brain development remain poorly understood, particularly in humans and closely related non-human primates. We describe a high resolution transcriptional atlas of rhesus monkey brain development that combines dense temporal sampling of prenatal and postnatal periods with fine anatomical parcellation of cortical and subcortical regions associated with human neuropsychiatric disease. Gene expression changes more rapidly before birth, both in progenitor cells and maturing neurons, and cortical layers and areas acquire adult-like molecular profiles surprisingly late postnatally. Disparate cell populations exhibit distinct developmental timing but also unexpected synchrony of processes underlying neural circuit construction including cell projection and adhesion. Candidate risk genes for neurodevelopmental disorders including primary microcephaly, autism spectrum disorder, intellectual disability, and schizophrenia show disease-specific spatiotemporal enrichment within developing neocortex. Human developmental expression trajectories are more similar to monkey than rodent, and approximately 9% of genes show human-specific regulation with evidence for prolonged maturation or neoteny.
The primate brain develops through a series of stereotyped processes that are conserved across mammals1, including the specification, migration and maturation of diverse cell types and the formation and refinement of functional neuronal circuits. There are also primate-specific features of brain development that increase anatomical2, cognitive3, and behavioral complexity4 and may explain why many human neurological and neuropsychiatric diseases are not well modeled in rodents5. These features include a larger cortical progenitor pool in the outer subventricular zone not present in rodents2 and protracted myelination6, synapse production7, and pruning8. Consequently, rhesus monkey and human share a greatly expanded neocortex and specialization of areas (most notably primary visual cortex2), compared to mouse, reflecting the more recent common ancestor of human and rhesus monkey (25 million years) than human and mouse (70 million years)9. Likewise, rhesus monkey and human brain10 share more similar patterns of gene expression than do mouse and human brain11. Rhesus monkey thus provides a valuable proxy for human brain development, particularly during prenatal and early postnatal development that are difficult to sample in human, and also provides a comparator to study human-specific features such as prolonged maturation or neoteny6,12,13.
Molecular programs responsible for brain development remain incompletely understood in any species, due to the spatiotemporal complexity of these processes and the resource intensity of methods needed to probe them. Transcriptome-based approaches have dramatically accelerated understanding of variation in gene expression programs related to brain structure and function in adult and developing human14–17 and rhesus monkey10,18, albeit with limited anatomical resolution and developmental coverage. Recent studies in adult mouse19,20 and human21 have profiled individual cortical cells and demonstrated robust transcriptional differences between neuronal and non-neuronal cell types and large-scale changes over the course of development. While these approaches offer cellular resolution in targeted domains, a global picture of gene expression changes over development across the complete cellular milieu has been lacking
This project aimed to create a transcriptional atlas of non-human primate brain combining fine anatomical precision with dense temporal coverage across prenatal and postnatal development in regions associated with neuropsychiatric disease. These transcriptional data are complemented by imaging and histology-based anatomical reference and cellular resolution in situ hybridization (ISH) datasets, all publicly accessible through the NIH Blueprint Non-Human Primate (NHP) Atlas (www.blueprintnhpatlas.org; www.brain-map.org).
Spatiotemporal transcriptome analysis
This resource characterizes rhesus monkey forebrain development with three modalities: 1) anatomical reference datasets consisting of magnetic resonance imaging (MRI) and corresponding densely sampled Nissl stains; 2) cellular resolution gene expression data generated with a high throughput ISH platform; and 3) fine anatomical resolution transcriptome timeseries data generated with a combination of laser microdissection (LMD) and DNA microarrays (Fig. 1a). Molecular analyses focused on discrete progenitor and postmitotic cell populations in neocortex (medial prefrontal and visual cortical areas), hippocampus, striatum, and amygdala (Fig. 1b, Supplementary Tables 1–2). Ten developmental timepoints were chosen to correspond to peak periods of neurogenesis for neurons destined for different layers and glial cell types of V1 (prenatal)22 and major developmental epochs: neonate, infant, juvenile, and young adult (postnatal; Fig. 1c). Anatomical parcellation at each time point was based on prior work in monkey and human brain development2,17 (Extended Data Fig. 1). Canonical genes marking different neural cell types showed expected spatiotemporal patterning across development, shown as heatmaps (Fig. 1d and Extended Data Fig. 2) laid out following the anatomical organization in Fig. 1c.
High-resolution transcriptional profiling of rhesus monkey brain development
To gain a high level understanding of transcriptional dynamics during cortical development, we represented the similarity between cortical samples using multidimensional scaling (MDS) based on correlated expression (Fig. 1 e–g). Layer and age explained almost 90% of the variation across all samples (n = 922; Fig. 1e) and prenatal samples (n = 542; Fig. 1f). To examine smaller differences between neuronal populations (grey circles in Fig. 1f), we repeated the MDS analysis only on V1 cortical layers (containing different types of excitatory neurons; n = 175; Fig. 1g). Transcriptional similarity of layers reflects spatial proximity and birth date, as described in adult rhesus monkey10. Moreover, continuous variation across prenatal development suggests gradual changes in gene expression.
There were few sex-related differences in expression, although the study was not powered to detect subtle differences restricted to specific brain regions or developmental stages. Only one autosomal gene (LOC693361, NADH-ubiquinone oxidoreductase chain 1-like) showed robust increased expression in males across brain regions during prenatal development (Extended Data Fig. 9, Supplementary Table 3). However, the microarray probe for LOC693361 also targets an unannotated region of rhesus monkey Y chromosome23 so there are likely no autosomal genes with persistent male- or female-specific expression differences across development.
Developmental transcriptional dynamics
Previous work has demonstrated that different cortical layers have distinct transcriptomic profiles both in adult and developing cortex, where the largest differences are seen between germinal and postmitotic cell populations17. In grossly dissected developing human prefrontal cortex, gene expression changes fastest prenatally and slows sharply just after birth14. If dividing cells drive most expression change, then depletion of the progenitor pool would explain decreased rates. In contrast, we find that all layers of V1 and prefrontal cortex (anterior cingulate gyrus; ACG) have similar rates of expression change across development (Fig. 2a). Thus transcriptional programs change dramatically both in progenitor cell populations as they give rise to different neuronal and non-neuronal cell types, as well as in postmitotic neuronal cell types as they differentiate and mature into adulthood.
Transcriptional dynamics across brain regions and ages
Similar developmental trajectories were seen in hippocampus, basal ganglia and amygdala (Extended Data Fig. 4a). However, the genes changing dynamically could differ by region or age. We identified the top 1000 genes with the largest increase and decrease in relative expression between all pairs of ages for each structure and visualized overlap of these gene lists as a heatmap (Fig. 2b). This threshold was selected because at least 1000 genes significantly changed expression (ANOVA FDR < 0.05, fold change > 1.25) between all adjacent ages in the majority of regions (Extended Data Fig. 4b). Genes changing over time were remarkably synchronized across different cell populations. At every age, approximately half of the top increasing and decreasing genes were shared by most brain regions, and unsupervised clustering grouped samples almost perfectly by age (age index bar in Fig. 2b). Genes increasing with age showed substantial overlap across ages within, but much less between, prenatal and postnatal ages. This suggests that distinct transcriptional programs are active in prenatal and postnatal development, and that many of these programs progress gradually. At early stages, samples grouped by proliferative state independently of brain region (mitotic index bar in Fig. 2b) reflecting a common set of cell division genes. Postnatally, samples grouped more by anatomical structure, reflecting increased regional identity during brain maturation (structure index bar in Fig. 2b).
To interpret these dynamic gene expression patterns, we searched for enrichment of gene ontology (GO) terms at each region and age (Supplementary Tables 4–5) and represented significant enrichments as heatmaps ordered as above (Fig. 2c). For example, genes related to onset (positive regulation) of axonal projection were enriched during prenatal periods as expected. Surprisingly, this process appears to be actively repressed during late postnatal development, as genes associated with offset (negative regulation) of axonal projection showed increased expression from juvenile (12 months) to young adult (48 month) stages. Processes associated with synapse development appear to be temporally enriched in late prenatal and early postnatal periods, and also synchronized between presynaptic (synaptic vesicle localization) and postsynaptic (dendrite development) neurons. Interestingly, the region-specific blocks of coordinated expression in postnatal visual cortex (V1 and V2) and hippocampus (Fig. 2b, arrows) were enriched for autophagy (Extended Data Fig. 4c), which may reflect early dendritic spine pruning24 and grey matter volume reduction25 in visual cortex.
Regional timing of biological processes
Some biological processes, such as synaptogenesis and myelination, are protracted over development6,24, and we find that process activation (increasing genes) is longer than inactivation (decreasing genes; Fig. 3a). Biological processes persist longer in all regions pooled together (black lines) than in any individual region (colored lines) so developmental timing must vary between regions. We quantified synchrony between pairs of regions as the proportion of ages during which a process is active in either region, for processes active in all regions at some age (e.g., Extended Data Fig. 5a). Remarkably, despite major differences in cellular and functional makeup of brain regions sampled, all pairs of regions showed similar degrees of synchrony across all GO terms. This synchrony was centered on overlap of approximately half of all ages (Extended Data Fig. 5b), far greater than expected by chance. Few processes – cell proliferation, projection, and adhesion – were synchronized across all regions and ages (Supplementary Table 5).
Variable onset of biological processes between brain regions
Processes usually started before birth in all regions, but sometimes at strikingly different times in different regions (Fig. 3b). Consistent with known early maturation of subcortical circuits1, rank ordering age of onset for each process across regions showed earlier subcortical onset, particularly in amygdala, and similar timing for neocortex and hippocampus (Extended Data Fig. 5c). For example, subcortical globus pallidus is myelinated before cortical white matter26 and shows early increased expression of genes specific to myelinating oligodendrocytes (MOG, MOBP, and ERMN) (Fig. 3c and Extended Data Fig. 5d). Interestingly, three other oligodendrocyte markers (MAL, ASPA, and OPALIN) turned on synchronously in all regions, suggesting a temporal and regional dissociation of different stages of myelination.
Late emergence of mature cortical identity
The neocortex is composed of diverse neuronal subtypes organized into layers that have distinct connectivity and transcriptional signatures in the adult10,20,27 with graded and discrete variation between cortical areas15,16,28. A neuron develops its mature molecular identity through cell-autonomous processes and environmental interactions during migration, process extension, and synapse formation. If early generated neurons closely resemble their mature states, then environmental interactions must have small effects on their terminal molecular identities. Since we found dramatic expression changes within postmitotic cortical layers (see Fig. 2a), there is a significant role for either or both of these mechanisms.
To examine the emergence of mature laminar signatures, we compared layer-enriched expression patterns across development from the earliest stage at which all postmitotic layers are present (E80) to adulthood. We calculated the average expression correlation, across layers and between all pairs of ages, of the top 2.5% of genes most specific for each layer in V1 at each age (see Methods, Supplementary Table 6). Adult identity emerged gradually (Fig. 4a), as evidenced by decreased expression similarity with increased age separation. In fact, laminar expression patterns in young adulthood barely resembled mid-fetal cortex (R < 0.1).
Protracted maturation of neocortex through young adulthood
These correlation trends result from nearly complete shifts in genes showing robust layer-specific expression (fold change > 1.5; Fig. 4b). Most genes were enriched only for a subset of contiguous ages at early, middle, and late stages, while few genes showed persistent laminar enrichment (Fig. 4b,c, Supplementary Table 7). Sets of laminar genes tile smoothly across development, suggesting a graded emergence of functional characteristics of neuronal subtypes as well as shifting proportions of subtypes within layers. Genes with early laminar expression are enriched for GO terms related to early developmental events such as neuron differentiation (P = 1.6 × 10−7) and axon generation (P = 6.7 × 10−6), whereas later laminar genes are enriched for axon guidance (P = 6.5 × 10−5) and innervation (P = 9.8 × 10−6).
Different anatomical and functional neocortical areas have distinct gene expression patterns in the adult, most prominently in V115,16,28. During development, intrinsic (protomap)29 and extrinsic (protocortex)30 factors shape cortical area identity. To ask whether the temporal dynamics of areal expression support intrinsic versus extrinsic areal specification, we first identified genes differentially expressed between caudal (V1, blue) and rostral (ACG, red) cortex in different layers at each developmental stage (Fig. 4d, Supplementary Table 8). More genes were differentially expressed between cortical areas in postmitotic layers during late postnatal than prenatal development. These genes were enriched for cell adhesion (P = 3.9 × 10−6) and synaptic transmission (P = 7.3 × 10−4) GO terms, suggesting a role for extrinsic factors related to the functional maturation of cortical circuitry. Developmental process timing varies across primate cortex, with neurogenesis occurring earlier in frontal than caudal cortex but circuit maturation persisting later in frontal cortex6,8, and this was recapitulated by an even more protracted emergence of enriched genes in ACG (12 to 48 months) compared to V1 (3 to 12 months).
To test for genes that may contribute to intrinsic specification, we searched for genes with persistent areal enrichment beginning prior to the arrival of thalamic31 or other extrinsic inputs. While 38% of genes showed regional differences, only 1% were persistent over development (Supplementary Table 8). These genes were primarily expressed in postmitotic cortical layers, for example CBLN217, and only 20 genes (0.2%) were expressed in germinal zones, including the well-known caudal-to-rostral gradient genes FGFR317,29 and NR2F1 (COUP-TFI)29. Thus, while there is evidence for a small number of genes clearly reflecting early intrinsic areal specification, there are many more adult differences that could be due to extrinsic interactions and activity-dependent mechanisms32.
Some areal differences are likely due to differences in the proportion and maturation state of specific cell types, such as dividing neural progenitors and postmitotic neurons and glia. Indeed, we identified spatiotemporal locations with significant enrichment in V1 or ACG for markers of cell cycle (i.e. actively dividing progenitor cells), GABAergic and glutamatergic neurons in adult mouse V1, and astrocytes (Fig. 4e–h). Cell cycle markers were enriched in V1 SZ from E80-E90 (Fig. 4e), reflecting extended generation of superficial excitatory neurons in V1 relative to ACG. Likewise, markers for excitatory neurons (Fig. 4g) were enriched in ACG prenatally, corresponding to earlier generation and maturation of these cells. Postnatal enrichment in V1 is likely secondary to using markers derived from adult mouse V1. Astrocyte marker enrichment in prenatal ACG progressed from proliferative to postmitotic layers (Fig. 4h), tracking astrocyte generation and migration in these layers. Finally, markers of inhibitory GABAergic interneurons showed early enrichment in V1 SZ, followed by ACG enrichment in several layers (Fig. 4f), a potential signature of early interneuron migration from medial ganglionic eminence to caudal cortex.
The majority of areal genes showed intermediate expression in primary somatosensory cortex (S1), located approximately midway between rostral ACG and caudal V1 (Extended Data Fig. 6). Therefore, rostro-caudal expression gradients, rather than specific features of ACG and V1, likely drive many of these areal expression differences.
Mapping disease genes to development
Highly heritable but genetically complex neurodevelopmental disorders will be more tractable if we can identify common pathways linking many candidate risk genes. For example, recent work on autism spectrum disorder (ASD) demonstrated that candidate genes are co-expressed in mid-fetal human cortical neurons17,33,34 and converge on synaptic development pathways. Large-scale genetic association studies have identified high confidence genes associated with primary autosomal recessive microcephaly (MCPH), ASD, intellectual disability (ID), and schizophrenia (SCZ). We localized these disease genes in developing cortex using this new high resolution map.
We used weighted gene co-expression network analysis (WGCNA) to identify sets of genes with correlated patterning (modules) in cortical samples at each age, linked modules with overlapping genes between adjacent ages, and annotated modules for enrichment in different cortical layers and cell types (Fig. 5a; Supplementary Table 9). Next, we tested disease gene sets for significant enrichment (hypergeometric test, corrected P < 0.1). As expected, MCPH genes showed early- to mid-fetal enrichment in neuronal progenitor-enriched modules (Fig. 5b) where they are positioned to alter cell cycle kinetics to reduce neuron numbers and brain size. Many ASD genes showed coordinated expression in post-mitotic neurons prenatally (Fig. 5c), as previously reported17,33,34, but also postnatally. In contrast, SCZ genes were selectively enriched only in postmitotic neuronal modules from infancy through adulthood (Fig. 5d).
Spatiotemporal localization of disease-specific associations in developing cortex
ID-associated genes were not enriched in any gene modules, consistent with the view that disruption of many biological pathways can undermine cognitive development34. To investigate whether these diverse molecular pathways converge on intermediate phenotypes within ID such as structural brain malformations, we looked for correlated expression among the 71 ID genes across prenatal and postnatal cortex. We identified four major expression patterns: 1) postnatal cortical plate, 2) prenatal germinal layers, 3) prenatal subventricular zone and cortical plate, and 4) more complex patterns spanning development (Fig. 5e). Consistent with the module analysis above, MCPH-associated genes had significantly similar expression profiles (multivariate distance matrix regression, P < 0.001) and were enriched in prenatal germinal zones in clusters 2 and 3 in either VZ or SZ. In contrast, other phenotypes did not share expression patterns, including cortical atrophy, seizures and even microcephaly with other brain malformations. Furthermore, genes with similar patterns can give rise to different phenotypes: ASXL3 and GRIP1 are co-expressed (R = 0.80) in cluster 3 but are associated with MCPH and corpus callosum agenesis, respectively.
Human-specific developmental trajectories
Humans exhibit unique cortical and cognitive developmental trajectories compared to other mammals5 that are likely correlated with differences in gene expression. To identify conserved and divergent genes, we compared gene expression trajectories across prenatal and postnatal development in frontal cortex between rat, rhesus monkey, and two independent human data sets, normalizing timescales by developmental event scores1. Expression conservation varied across species (Kruskal-Wallis rank sum test, P < 10−21), with greater similarity within humans than within primates (Wilcoxon signed rank test, P < 10−6), as expected based on evolutionary distance (Fig. 6a). Surprisingly, there was no significant difference in conservation of rat trajectories with either human or rhesus monkey, indicating transcriptional patterning has evolved at the same average rate in human and rhesus monkey since our last common ancestor. On average, 69% of genes had conserved expression trajectories (R > 0.5) across all three species, and genes with the most dynamic expression change over development were even more highly conserved (Extended Data Fig. 7a,b). For example, EMX2, a transcription factor critical to cortical arealization35, and CNTN1, a cell adhesion gene that guides neuronal migration and connectivity (Fig. 6b), are highly conserved.
Conserved and human-specific gene expression trajectories in frontal cortex
A significant minority (13%) of genes differed between primates and rat (validated for a subset of genes in developing mouse; Supplementary Table 10) and approximately 9% had human-specific expression trajectories. The proportion of genes showing different conservation patterns was robust to gene selection and correlation threshold (Extended Data Fig. 7b). Some of these differences were dramatic; for example, BMP3 decreases over primate development but increases in rodents (Fig. 6b). BMP3 is a WNT signaling growth factor that has been linked to craniofacial variation in dogs36. CNTN2, a close family member of the conserved adhesion gene CNTN1, likewise shows opposite developmental trajectories in primates and rodents. In both human data sets, LGALS1 and LIN7A increase or decrease expression, respectively, in opposite directions to other species. LGALS1 further has a different adult laminar pattern of cortical expression in human compared to mouse11, while deletion of LIN7A is associated with human intellectual disability and disrupted neuronal migration and axonal pathfinding in mouse37.
Some aspects of cortical development, such as myelination6 and synaptic pruning8, are protracted or neotenous between cortical areas or between species. Reasoning that abrupt changes in expression may represent important developmental milestones, we compared developmental timing by identifying distinct changes in expression trajectories, or breakpoints. 179 increasing and 179 decreasing genes met these criteria in all three species, being well fit by segmented linear regression (Supplementary Table 11). Consistent with different developmental rates of these species, breakpoints for increasing (Fig. 6c) and decreasing (Extended Data Fig. 7c) genes often occurred earliest in rat, intermediate for rhesus monkey, and latest in human. Almost half (81) of increasing genes were synapse related (dashed lines, Fig. 6c), and breakpoints coincided with ages of peak synaptic density estimated in these species (shaded rectangles, Fig. 6c, Extended Data Fig. 7d). There was no significant difference in the breakpoints for synapse-related genes in human V1 versus prefrontal cortex and only a short delay in rhesus monkey V1 (42 days, P = 9.6 × 10−9, Extended Data Fig. 7e), consistent with synchronous synaptogenesis reported in primates7,38–40. One study later reported protracted synaptogenesis in human41, but this study suffered from sparse sampling and no statistical analysis and does not appear consistent with the bulk of evidence.
While breakpoint timing was largely conserved between species, particularly between rhesus monkey and rat (Extended Data Fig. 7f), human breakpoints were clearly bimodally distributed (Fig. 6c) suggesting that some genes had different developmental timing. For example, OLIG1 is expressed in maturing oligodendrocytes and had a late breakpoint relative to rhesus monkey (Fig. 6d), consistent with prolonged myelination in several human cortical areas compared to rhesus monkey and chimpanzee6. Surprisingly, many more genes had an early rather than late breakpoint in human prefrontal cortex (Fig. 6e), in contrast to previous work12,13 showing human-specific delayed peak expression in this brain region. However, we found that many early genes continued to increase expression through adulthood (Fig. 6e), such as SYT7 (Fig. 6d), a presynaptic calcium-binding protein important for modifying neurotransmitter release42. In rhesus monkey, gene expression often did not change significantly after breakpoints. Genes with an early breakpoint and late maximal expression may mark developmental processes that are protracted in human relative to non-human primates and that underlie our extended cognitive development. Interestingly, we found this transcriptional signature of neoteny between species in both prefrontal cortex and V1, but much less so in evolutionarily older structures including hippocampus, amygdala and striatum (Fig. 6e and Extended Data Fig. 7g).
The current project transcriptionally profiled primate brain development with fine anatomical detail from early gestation to young adulthood. While expression rates of change decreased more than 100-fold over development, we observed small but coherent changes from juvenile to adult, a period of enormous cognitive change and susceptibility to neuropsychiatric disease. These changes related to the late emergence of mature laminar and areal signatures, biological pathways such as negative regulation of axonal pathfinding, and the appearance of gene modules significantly associated with ASD and SCZ.
Two unexpected characteristics of expression trajectories support extrinsic influences on developmental transcriptional regulation. First, there was a striking synchrony among genes changing in disparate brain regions and cell populations, suggesting a mechanism for global regulation such as circulating hormones. Second, the surprisingly late acquisition of adult-like cortical areal and laminar molecular phenotypes points to an important role for contextual and activity-dependent mechanisms in sculpting mature cellular phenotypes43,44.
We show that 22% of genes have different developmental trajectories in rat and human, comparable to 25% of genes that have different laminar patterns in adult mouse and human cortex11. In contrast, approximately 9% of genes have different trajectories in rhesus monkey and human including genes with delayed peak expression solely in human cortex. Therefore, rhesus monkey is an important comparator for understanding human-specific features of brain development but cannot fully model protracted circuit formation and associated diseases that are seen only in humans.
A recently expanded set of ASD candidate genes45 were significantly enriched in newborn neurons during prenatal development, as previously reported17,33,34, but also enriched throughout postnatal development. SCZ risk genes were also enriched in neurons but not until infancy, suggesting a larger role for dysfunction in circuit refinement than prenatal processes such as neurogenesis. SCZ and ASD were enriched in the same neuron-enriched module in infancy despite only 5% overlap in candidate genes, pointing to a shared etiology that is being increasingly appreciated46.
This data resource (www.blueprintnhpatlas.org; www.brain-map.org) has many potential applications; for example, establishing an in vivo baseline against which to compare the identity and maturity of in vitro stem cell-derived neurons47. Recent technical advances enable profiling the full transcriptome19–21, epigenome48, and ultimately proteome of single cells. These techniques promise to refine this broad survey to a causal understanding of molecular programs driving the complete lineage of primate brain cells and the maturation of specific neuron types in functional circuits.
No statistical methods were used to predetermine sample size, and the experiments were not randomized or blinded. More complete descriptions of the experimental and data processing methods are available in white papers at the NIH Blueprint NHP Atlas website (www.blueprintnhpatlas.org; “Documentation” tab).
Frozen postmortem tissue samples from prenatal rhesus macaque (Macaca mulatta) of predominantly Indian origin (Extended Data Table 1) were provided by the Time-Mated Breeding Program at the California National Research Primate Center (CNPRC; www.cnprc.ucdavis.edu). Prenatal brain material was acquired following fetectomy using Standard Operating Procedures (SOPs) at the CNPRC. Timed pregnancy derived biological replicate specimens (2 males, 2 females) at each of six prenatal developmental stages (E40, E50, E70, E80, E90, and E120) were profiled. These timepoints were selected to coincide with peak periods of neurogenesis for the different layers of primary visual cortex based on birthdating experiments22. Frozen postmortem tissue samples from postnatal rhesus macaque were also provided by CNPRC. Brain regions were systematically collected from well-characterized rhesus monkeys born and raised at the CNPRC in outdoor, half-acre enclosures that provide a naturalistic setting and normal social environment. For transcriptome analysis, three male specimens at each of four postnatal developmental stages representing the neonate (0 months), infant (3 months), juvenile (12 months) and post-pubertal adult (48 months) were profiled. Extensive health, family lineage and dominance information were maintained on all animals. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at UC Davis.
Postnatal MRI and histological reference series
An interleaved anatomical atlas of magnetic resonance imaging (MRI), histology, and block face photographs was created from four male rhesus macaques at 2 weeks, 3, 12 and 48 months after birth as described in the “Reference Series” white paper. In addition, ISH data were generated serially across complete hemispheres of three post-pubertal adult (48 months) male specimens. Genes were selected to mark specific cell populations and cortical areas and to represent gene families important to neural function and associated with neuropsychiatric disease and brain evolution. Several histological stains (Nissl, acetylcholinesterase and SMI-32) of complete hemispheres were combined with ISH data to provide an unprecedented cellular resolution view of gene expression in the whole post-pubertal adult brain. High-throughput colorimetric ISH methods are described in detail elsewhere49 and in the “In Situ Hybridization” white paper.
Following collection, brain tissues were partitioned in a manner dependent on specimen stage, gradually frozen, and then stored at −80°C until processing. For a subset of E40 specimens, following specimen collection, the calvarium was frozen intact in an isopentane/dry ice slurry maintained at −40°C to −45°C, gradual freezing at a steady rate. For the majority of E40 specimens, the specimens were embedded in OCT (optimal cutting temperature compound) during the freezing process. In brief, chilled OCT was placed around the calvarium. A disposable embedding chamber was filled with approximately 5 mm3 chilled OCT. The specimen was carefully oriented and centered in the OCT, posterior surface down in the OCT. Then, the specimen was aligned along the medial/lateral axis using the bilateral ocular fiduciaries as a frame of reference. Next, the specimen was aligned in the coronal plane. After alignment along all axes, OCT was added to encase the specimen in its entirety. The top of the specimen was covered with approximately 3 mm of OCT. The chamber containing the specimen was directly placed onto a level bed of dry ice. The specimen and OCT were allowed to freeze completely. After demarcation of the orientation of the brain in the OCT block, the frozen tissue block was stored at −80°C.
For the E50 specimens, following removal of the brain from the skull, the whole brain was frozen intact in an isopentane/dry ice slurry maintained at −40°C to −45°C, gradual freezing at a steady rate. For a subset of E50 specimens, the specimens were embedded in OCT during the freezing process. In brief, chilled OCT was placed around the intact brain. Freestanding aluminum bars were assembled onto a Teflon coated plate and sized to the appropriate specifications for the E50 brain. The internal chamber was filled with approximately 5 mm3 chilled OCT. The specimen was carefully oriented and centered in the OCT dorsal surface down in the OCT. Then, the specimen was aligned along the medial/lateral axis using the longitudinal fissure as the frame of reference. Next, the specimen was aligned in the coronal plane. After alignment along all axes, OCT was added to encase the specimen in its entirety. The top of the specimen was covered with approximately 3 mm of OCT. The Teflon plate containing the specimen was directly placed onto the level bed of dry ice. The specimen and OCT were allowed to freeze completely. After demarcation of the orientation of the brain in the OCT block, the aluminum bars were removed and the frozen tissue block was stored at −80°C.
For the E70, E80, E90, and E120 specimens, the hemispheres were bisected along the midline and individually frozen by placing the medial aspect of each hemisphere down onto an aluminum-Teflon coated plate that was slowly lowered into an isopentane/dry ice slurry maintained at −40°C to −45°C. Only approximately a third of the tissue was submerged in the slurry to allow the tissue to gradually freeze and to keep freezing artifacts to a minimum. Frozen hemispheres were stored at −80°C.
Depending on the prenatal timepoint, different approaches were taken for generating coronal slabs. When possible, the number of slabs per specimen was kept to a minimum. The E40, E50, and E70 specimens were not slabbed. For E80, the first slab contained up through the temporal pole and the second slab contained the occipital pole. For E90, the first slab contained the frontal lobe anterior of the temporal pole and the second slab contained temporal pole posterior through the occipital lobe. For E120, three coronal slabs were made. The first slab consisted of the frontal lobe anterior of the temporal pole. The second slab consisted of the temporal pole posterior to the cerebellum including all of the mid-brain. The third slab included primarily the occipital lobe.
For postnatal brains, after dissection brains were sectioned into coronal slabs approximately 1 to 1.5 cm in thickness and the left hemisphere was prepared for sectioning onto microscope slides for ISH. Structures for microarray analysis were isolated from the right hemisphere slabs, and these samples were then frozen at −80°C until processed further.
Laser microdissection and RNA isolation
Tissue spanning five anatomically distinct brain regions — frontal cortex, visual cortex, hippocampus, striatum, and amygdala (Figure 1; Supplementary Table 1) — was selected from each specimen and processed for further thin sectioning and LMD using a standard protocol. Specifically, frozen tissue was cryosectioned at 14 µm onto polyethylene naphthalate (PEN) slides (Leica Microsystems, Inc., Bannockburn, IL) and a 1:10 Nissl series was generated for neuroanatomical reference for all prenatal timepoints. In addition, for the E40, E50, E70, E80, and E90 timepoints, a 1:10 GAP43 and 1:10 ENC1 in situ hybridization (ISH) series was generated for neuroanatomical reference, as they often clearly delineate nuclei and layers at early developmental stages. For E120, a 1:10 acetylcholinesterase series was generated for neuroanatomical reference. In total 127 transient and terminal anatomical structures were isolated using this strategy50–53.
After drying for 30 minutes at room temperature, PEN slides were frozen at −80°C. Slides were lightly Nissl stained with cresyl violet to allow cytoarchitectural visualization. Slides were fixed in ice-cold 70% ethanol for 30 seconds, washed 15 seconds in nuclease-free water, stained in 0.7% cresyl violet in 0.05% NaOAc, pH 3.4 for 2 minutes, nuclease-free water for 15 seconds, followed by 15 seconds each in 50%, 75%, and 95% ethanol, followed by 20 seconds in 100% ethanol, and then a final 100% ethanol wash for 25 seconds. Slides were air-dried for 2 minutes, and desiccated in a vacuum for 1 hour, then frozen at −80°C until microdissection. Laser microdissection was performed on a Leica LMD6000 (Leica Microsystems, Inc., Bannockburn, IL), using the Nissl stain and GAP43 and ENC1 ISH or acetylcholinesterase histological staining as a guide to identify target brain regions in prenatal samples.
Microdissected tissue was collected directly into RLT buffer from the RNeasy Micro kit (Qiagen Inc., Valencia, CA) with β-mercaptoethanol. Samples were volume-adjusted with water to 75 µl, vortexed, centrifuged, and frozen at −80°C.
RNA was isolated for each structure following the manufacturer’s directions for the RNeasy Micro kit. RNA samples were eluted in 14 µl and 1 µl was run on the Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA) using the Pico assay. Due to low sample volume and incompatibility of the eluant with the Nanodrop spectrophotometer (Thermo Scientific, Wilmington, DE), samples were quantitated using the Bioanalyzer concentration output. This was done by running a 1ng/µl RNA standard on the same Pico chip and then dividing the sample concentration output by the output of the standard concentration. For prenatal samples, 2 ng of total RNA was almost always used as the input amount for the labeling reaction, and the average RNA Integrity Number (RIN) of passed samples was 7.5, with RINs typically lower than 4.5 failing. For postnatal samples, 5 ng of total RNA was almost always used as the input, and the average RIN of passed samples was 7.5, with RINs typically lower than 5 failing.
For the Nissl neuroanatomical reference slides, slides were stored at 37°C for 1–5 days prior to staining. Sections were defatted with xylene or the xylene substitute Formula 83, and hydrated through a graded series containing 100%, 95%, 70%, and 50% ethanol. After incubation in water, the sections were stained with 0.213% thionin, then differentiated and dehydrated in water and a graded series containing 50%, 70%, 95% and 100% ethanol. Finally, slides were incubated in xylene or Formula 83, and coverslipped with the mounting agent DPX. After drying, slides were cleaned prior to digital imaging.
UNIVERSITY OF SOUTH AUSTRALIAOpen Universities Australia
ASSIGNMENT COVER SHEET
Assignment cover sheet
needs to be included at the front of each assignment.
Please fill in all details clearly
Andrew Edward Cole
PO Box 9043Deakin LPOACT
UniSA student ID
1 1 0 0 2 0 0 71
0407 384 998
Accounting, Decisions and Accountability
Due date (indicate if extension granted):
May, 2009 (1 Week Extension Granted)
2Assignment topic as stated in the
I declare that the work contained in this assignment is my own, except whereacknowledgement of sources is made.I authorise the University to test any work submitted by me, using text comparison software,for instances of plagiarism. I understand this will involve the University or its contractorcopying my work and storing it on a database to be used in future to test work submitted byothers.I understand that I can obtain further information on this matter athttp://www.unisanet.unisa.edu.au/learningconnection/student/studying/integrity.aspSigned:
Andrew Cole – 13 May 2009
Note: The attachment of this statement on any electronically submitted assignments will bedeemed to have the same authority as a signed statement.
Date received fromstudent
Academic staff memberRecorded anddispatched