Decoding human gene expression signatures in the brain

Abstract

The evolution of higher cognitive functions in humans is thought to be due, at least in part, to the molecular evolution of gene expression patterns specific to the human brain. In this article, we explore recent and past findings using comparative genomics in human and non-human primate brain to identify these novel human patterns. We suggest additional directions and lines of experimentation that should be taken to improve our understanding of these changes on the human lineage. Finally, we attempt to put into context these genomic changes with biological phenotypes and diseases in humans.

Keywords: :

• gene expression
• genomics
• coexpression
• brain
• nervous system
• evolution

Introduction

Identifying evolutionary changes that have contributed to the increased cognitive capacities of humans has presented unique challenges. Almost 40 y ago, King and Wilson demonstrated that human and chimpanzee proteins differed by very little and hypothesized that changes driving human molecular evolution may stem from regulatory element divergence leading to changes in gene expression.¹ When the human and chimpanzee genomes were sequenced 30 y later,² this hypothesis turned out to be extremely prescient. In coding regions, the genomes were highly identical, but recent studies have suggested that there is significant divergence in non-coding and regulatory regions. In particular, there is evidence that segmental duplications and copy number variations have evolved on the human lineage and many of these regions are hotspots for regulatory factors such as transcription factors.^3-6 Thus, evolutionary changes in non-coding areas of the genome could have a significant functional impact on expression and regulation of expression. Recent studies such as the ENCODE project have identified up to 80% of the human genome as being transcribed.⁷ However, the function of the great majority of these transcripts is unknown, and whether these transcripts are conserved among primates is also largely unknown. Therefore, a large percentage of the genomic landscape remains unexplored in terms of functional output in the brain, or any human tissue. Finally, there is evidence for accelerated evolution of specific transcription factors such as FOXP2,⁸ and ZNF (zinc finger) transcription factors,⁶ among others. Together, these findings suggest that we have only begun to scratch the surface in terms of our understanding of how human genomic evolution has yielded important functional consequences in human phenotypes such as higher cognition.

Gene expression divergence among species

High-throughput technologies such as microarrays and next-generation sequencing (NGS) have allowed for the field of comparative genomics to blossom, especially with regards to comparative brain gene expression. These studies have the lofty goal of attempting to understand human-specific features such as language and higher cognition. As genetic and invasive manipulations in humans (and chimpanzees) are not ethically possible, these comparative transcriptomic studies from post-mortem brain tissue are an essential first step. In addition to potentially providing insight into human cognitive features such as language, one other driving force behind these studies is to gain a greater understanding of cognitive disorders such as schizophrenia and autism. As previously put forth,⁹ vulnerability to cognitive diseases such as schizophrenia likely evolved as a consequence of the cognitively developed human brain. This hypothesis is challenging to prove, as we cannot be completely certain that non-human primates do not develop cognitive disorders such as schizophrenia, for example. It is also somewhat of a circular argument in that disorders of cognition like schizophrenia and autism have a language component, yet language is thought to be a human-specific feature. Nonetheless, comparative primate brain gene expression studies have identified an enrichment of genes involved in these cognitive disorders expressed in human-specific patterns,¹⁰ supporting the validity of this hypothesis.

Shortly after the initial drafts of the human and chimpanzee genomes confirmed the small number of coding differences at a genome-wide scale, a number of investigations were undertaken to identify functional genomic differences between them on other levels. As microarrays were the ideal technology for providing genome-wide expression information,^11-13 they were employed to detect transcriptomic differences between human and non-human primates. One of the first studies using blood, liver, and brain reported that human brain has an accelerated number of gene expression changes compared with chimpanzees, orangutans, and macaques.¹⁴ An additional interesting finding was that there are more genes being upregulated on the human lineage.¹⁵ These data are intriguing in light of recent studies showing upregulation of expression in human fetal brain in genes defined as having a recent evolutionary origin in primates (‘young genes”).¹⁶ The importance of this increase in upregulated genes remains unknown but speculatively those genes should play more important roles due to their elevated expression during evolution. Among these young genes are the genes encoding ZNF proteins. One way to rapidly change a system such as the brain transcriptome would be to modify regulatory genes such as the ZNF transcription factors. Thus, even small increases in these genes could lead to significant downstream consequences. Thus, it may not necessarily be true that it is the increase in the genes themselves that are advantageous evolutionarily but rather that the resultant modification to the network overall is the important change. Increases in other classes of genes, such as those important for myelination, may have assisted in the evolution of the human brain having greater and prolonged periods of myelination and as a consequence increased processing power and speed.^17-19 Of course, downregulation of other groups of genes can also have an important impact. For example, since the human brain is thought to have greater plasticity important for enhanced learning and cognition, downregulation of genes important for neuronal differentiation or dendritic spine maturity may result in a more plastic brain milieu.

Although microarrays have proven to be a powerful tool for obtaining genome wide expression information to identify differences among primates, they have some disadvantages germane to our discussion here. For example, a significant percentage of data must be discarded when applying chimpanzee samples to human microarrays due to potential imperfect sequence matches, In addition, for direct comparisons of any two species-specific microarrays care must be taken to account for potential differential hybridization affinities of species-specific probes. Moreover, if the sequencing depth is sufficiently high, many types of transcripts that could not be detected by microarrays without specialized arrays can now be detected and quantitated such as non-coding RNAs, microRNAs and isoform information using RNA-seq. Our recent analysis shows that RNA-seq identified 25–60% more expressed genes than microarray and five to eight times more differential expressed genes between a number of human and chimpanzee brain regions.¹⁰ Thus, by using this newer technology, more detailed information about gene expression differences among human and non-human primates are being revealed.

Gene expression divergence among brain regions

Genome-wide transcriptional data can also be used to detect regional differences in the brain. Comparisons of multiple cortical areas in humans, chimpanzees and rhesus macaque supported the idea of gene expression upregulation on the human lineage.²⁰ In addition, comparative analysis among four cortical areas, caudate nucleus, and cerebellum in human and chimpanzee brains also revealed that these areas of the brain have a significant number of gene expression differences as might be expected.²¹ However, this study interestingly found few cortical regional differences within individuals, suggesting that modifications to gene expression in specific brain areas may not be sufficient to explain the functional differences among cortical areas. One of the most revealing comparative studies in frontal cortex and cerebellum of humans, chimpanzees and rhesus macaques included samples from throughout brain development.²² This study noted a striking shift in gene expression patterns of key neuronal genes to later in development specifically in the human data set. These data correlate well with known patterns of neuronal development and behavior in humans. As noted below however, additional brain areas and developmental time points are still needed to fully flesh out the extent to which regional and developmental expression patterns have been important for cognitive changes. In particular, brain regions that may be particularly vulnerable in cognitive disorders (anterior cingulated cortex or other frontal cortical regions) and the evolution of language (Broca’s and Wernicke’s areas) need to be investigated in more detail. How the developmental time periods link to disorders also needs to be better understood. For example, schizophrenia and autism are developmental disorders in which it is thought that fetal brain development may be the critical time period at risk for these diseases. While it is unlikely that a fetal brain expression study in any of the great apes will ever be performed, developmental brain expression has already been assessed in humans,^23,24 and could certainly be undertaken in a variety of monkey and other mammalian species.

Splicing differences

Other comparative genomics studies in the brain have focused on splicing as a mechanism for differential gene expression in primates. Human-chimpanzee comparisons suggested that differentially splicing affects a group of genes separate from those that are otherwise differentially expressed.²⁵ Further splicing analysis using microarrays that included rhesus macaque as an outgroup narrowed the number of human-specific brain splicing events.²⁶ More recent studies using RNA-seq in life-span data sets from multiple brain regions have expanded the data on human-specific splicing even further demonstrating a pervasiveness to alternative splicing in the differential regulation of gene expression in the human brain.²⁷ We have also shown that human-specific patterns of gene expression in the frontal pole occur at the exon level and demonstrate that one particular human-specific network is enriched for binding sites for the RNA-binding factor ELAVL2.¹⁰ Not only can ELAVL2 regulate gene expression through splicing but it has also been known to associate with microRNAs to regulate gene expression.²⁸ Thus, splicing that specifically occurs on the human lineage can have widespread effects on signaling pathways that may be critical for human brain function and cognition. Additional studies that link these splicing changes to overall gene expression changes in a more detailed spatio-temporal manner will be needed to more fully appreciate the role of splicing in human brain development and evolution.

Transcription factor/miRNA differences

Arguably, the most unique advanced cognitive feature of the human brain is the capacity for language, and a number of studies have attempted to identify molecular signatures of language by examining gene expression and regulation.¹³ FOXP2 is a transcription factor that when modified results in deficits in human speech and language.^29-31 FOXP2 has undergone rapid modification in the primate lineage including two amino acids changes specific to humans and Neanderthals^8,32,33 and noncoding promoter and intronic changes specific to humans.³⁴ By using ChIP-chip, FOXP2 target genes were identified in fetal human brain, and interestingly, many of these targets are either under positive selection in the human genome or differently expressed between human and chimpanzee brain.³⁵ A more recent analysis also determined that FOXP2 target genes from several studies are under positive selection.³⁶ Together, these data suggest that FOXP2 and its target genes may have co-evolved to regulate language-related transcriptional networks. Further experiments showed that changing the two human-modified amino acids could lead to changes in gene expression in human cells and mice as well as changes in rodent behavior.^37,38 Another example of human-specific gene regulation was found when we examined whether human FOXP2 differentially regulated gene expression compared with chimpanzee FOXP2.³⁷ Indeed, differentially regulated genes and co-expression networks were identified and were enriched for genes involved in brain development supporting the role of FOXP2 in the development of neural circuits related to the most salient of human-specific phenotypes, language. These data become even more relevant when we uncovered a module of gene co-expression specific to the human frontal pole that contained FOXP2, FOXP2 target genes, and an enrichment of genes involved in neuronal process formation.¹⁰

The focus on the role of FOXP2 in language evolution is a bit misleading. It is extremely unlikely that FOXP2 and its downstream target genes are the only genes involved in language. We expect that there will be many more genes identified in future studies that can be directly linked to language. However, it will be just as challenging to demonstrate that these new genes have no connection to FOXP2 at all. It is very possible that every expressed gene in the human brain feeds back or forward to every other expressed gene in some infinitely complex manner that we have not begun to appreciate. Moreover, FOXP2 should not be discussed only in light of its role in language. FOXP2 is a highly conserved protein that certainly plays an important role in other species that do not have language as it is defined in humans. FOXP2 is important for sensory-motor integration and at the cellular level has a function in neurite outgrowth and neuronal plasticity.³⁹ Thus, there must be other proteins with important conserved roles in brain development that have been modified in subtle ways to promote the evolution of language.

Other recent studies used a combination of microarrays and RNA-seq to compare between humans and macaques using prefrontal cortex tissue over the life span and identified a regulatory role of transcription factors and miRNAs in the post-developmental stage and differences in the conservation of gene expression between developmental and aging periods.⁴⁰ More detailed analysis of these data revealed that the expression of miRNAs in the brain is widely divergent on the human lineage compared with both chimpanzee and rhesus macaque brains.⁴¹ The miRNAs with human specific expression are frequently expressed in human neurons and tend to regulate genes related to neuronal function.⁴¹ Taken together, these data demonstrate the increased power of RNA-seq to identify small non-coding RNAs that play an important regulatory role in neuronal physiology.

Differences in co-expression networks

The above highlighted selection of comparative gene expression data emphasize the importance of upregulation of gene expression in the evolved human brain¹² and suggest that gene expression changes specifically in the cortex was critical for human brain evolution.⁴² However, straightforward differential expression analyses are not sufficient to capture all of the complexities and relationships in large-scale gene expression data sets. A number of studies have uncovered that the pattern of gene co-expression is a critical mediator of brain function and exhibits significantly different patterns across species.⁴³ Gene pairs that have strong co-expression relationships are more likely to be functional together; therefore such analysis of co-expression networks in the brain among primates has proven informative. For example, weighted gene co-expression network analysis has been used to identify human-specific networks of gene expression compared with chimpanzees.⁴⁴ It has also been used to identify cell type related modules in the brain that can be queried in future gene expression studies across primates⁴⁵ and to identify network modules that are related to human-specific FOXP2 regulation of gene expression.³⁷ Other similar network methods such as weighted topological overlap have been employed to identify human-specific links to transcription factors compared with chimpanzee.⁴⁶ It is likely that the combination of these and other methods that have been used to uncover human modifications as well as new algorithms will be necessary to fully mine all of the ongoing data accumulation.

In our recent study comparing gene expression in three brain regions from humans, chimpanzees and rhesus macaques,¹⁰ we identified human-specific co-expression networks. In particular, those networks that were specific to the frontal pole are of evolutionary importance as the frontal pole has undergone expansion and modification on the human lineage.^47,48 Therefore, the highly connected genes in these human-specific frontal pole networks are likely important players in the evolution of human cognitive functions. Future studies that investigate the functional consequences of these genes as discussed below should be informative.

Future directions and additional considerations

Although tools such as NGS are powerful for obtaining genome wide transcriptional signals and the analysis pipelines are becoming increasingly mature, there are still several potential technical problems that need to be addressed in more detail such as RNA degradation in the samples,⁴⁹ GC content bias in the sequencing data⁵⁰ and 3′/5′ bias.^51,52 All of these examples of potential technical challenges are particularly pertinent to the study of comparative genomics in primate brains as these samples are derived from post-mortem brain tissue which can often have a wide range of quality metrics including post-mortem interval, pH, cause of death, and sample storage. In addition, just like the challenges in comparing hybridization biases between species, the differential genomic features between species (like GC content) may indirectly affect measurements such as transcript abundance. Fortunately, an increasing number of tools are available to solve these issues^51,53-55 and more detailed pipelines on the analysis of such data features are available as well.^52,56-58 Thus, future comparative genomics studies in the brain need to be cognizant of these technical issues and apply appropriate tools to address them.

One major caveat to most of these studies is the inability to take into account how the environment or epigenetics may be affecting brain gene expression. As DNA methylation can potentially have a significant impact on gene expression, we have investigated the correlation of gene expression differences between humans and chimpanzees in the frontal pole to differential methyl-seq data from the same brain region.⁵⁹ Hundreds of hypomethylated genes in human frontal pole compared with chimpanzee were identified. Strikingly, there was a strong correlation between methylation status and gene expression suggesting that an important evolutionary mechanism for the regulation of gene expression in the human brain is due to promoter methylation status. Of course, other mechanisms have almost certainly evolved to modulate gene expression specifically in the human brain. For example, histone methylation could specifically be differentially regulated across primate brains. Work using lymphoblasts and prefrontal cortical neurons have already identified differential methylation in primates;^60,61 however, future studies using additional brain regions should uncover any tissue-specific changes in the human lineage. These identified changes in methylation, however, may obviously stem from differences among primates in terms of natural environment: humans have a rich and diverse environment with unfettered access to a variety of nutritional sources while the other primates studied are either living in research facilities or zoos where they have a limited environmental repertoire and a regimented diet. Recent longevity studies in primates have determined that different types of diets and potentially genetic background can impact conclusions about caloric restriction and lifespan.⁶² Thus, the importance of the interplay of evolved genetics and environment cannot be undersold. While it is unlikely that we will ever be able to fully control for environment in comparative primate studies, experiments in which identified expression and regulatory changes are modified in an environmentally controlled model system such as in a rodent should help to tease apart these interactions.

Another critical direction for identifying human-specific gene expression changes includes extending comparative analyses to many more brain regions and over development when possible as has recently been done in human brain,^23,24,63 and to a more limited extent in rhesus macaque.⁶⁴ For studies of the neocortex it will also be critical to capture layer-specific information as was done in macaque.⁶⁴ Such layer- and region-specific data will be insightful for cognition as many disorders such as autism and schizophrenia display cortical connectivity issues.^65-68 Thus, differences among primates in expression among cortical areas or upper cortical layers, which contain neurons that project intracortically, might uncover human vulnerabilities for these disorders. These studies will ultimately be challenged by the paucity of non-human primate tissue as well as the issue of tissue heterogeneity. Even within a given species the pieces of brain tissue that are used for transcriptome profiling contain potentially hundreds of types of cells. Therefore, as sequencing technologies are able to handle even smaller amounts of starting material the use of cell-specific approaches such as laser dissection and antibody-enriched cell sorting will become more feasible. Previous work has demonstrated that neuronal nuclei can be sorted from post-mortem brain tissue in humans, chimpanzees and macaques and used for comparative histone methylation studies.⁶¹ In silico approaches can also be used to bioinformatically dissect cell-type specific modules of gene co-expression.⁴⁵ And obviously moving beyond gene expression to differential proteomics should provide significant insight into the functional consequences of transcriptional evolutionary modifications. In particular, examination of the proteins that are involved in spines or, specifically, the postsynaptic density complex—which is potentially the most complex protein complex in neurons—^69,70 across primates species should prove informative into increased plasticity and learning in humans. A number of studies have already ascertained differential protein profiles in primate brains,^27,40 and future studies with additional brain regions and improved protein detection technology will ultimately expand upon these findings.

Finally, incorporation of imaging data such as fMRI from humans and DTI (diffusion tensor imaging) from a variety of primates (both post-mortem and in vivo) together with genotype and gene expression (post-mortem) information will be a novel methodology for combining the few rare data sets that are able to be collected from primate brains. Such genotype-phenotype studies should yield unprecedented insight into how gene expression may be driving brain structure and activity. However, these studies will be significantly challenged by the inability to directly manipulate gene expression in great apes and study the resultant phenotypic changes. Nevertheless, creative approaches in rodents (or monkeys) may take advantage of new genome editing techniques such as zinc finger nucleases,⁷¹ TALENs,⁷² and CRISPR⁷³ to rapidly generate animals with multiple genetic manipulations reminiscent of human-specific brain gene expression and regulatory patterns. These modified animals may then be studied in detail for behavioral modifications, drug responsiveness, and brain patterning and activity. Already, genetic comparisons have identified a human-specific gene duplication in SRGAP2 that leads to neuronal spine modifications that can be phenocopied using in vivo manipulations of SRGAP2 expression in mice.⁷⁴ Another recent study demonstrated that vertebrate-specific duplication of the Dlg gene led to alterations in cognitive function,⁷⁵ which is fascinating given the changes in expression in Dlg gene family members specifically in the human brain.^10,76 Thus, translating comparative genetics/genomics studies into new animal models has already proven to be insightful into how human brain evolution has promoted increased cognitive abilities. We cannot of course discount the importance of using cellular models, in particular human cell models, as we and others have done for studying the evolution of FOXP2 function, for example.³⁷ Neither transgenic animals nor human cellular models will be sufficient on their own though to truly provide insight into how human genetic modifications affect cognition. However, these are the only options available. If behavioral assays in animals are interpreted cautiously and conservatively and these data are combined with functional studies in human cells and confirmation studies in primate brain tissue (in situ hybridizations, immunohistochemistry, etc.), only then will we begin to link the evolution of genes to behavioral outputs and disease manifestations. In summary, the combination of whole genome sequencing of specific patient populations with cognitive impairments, brain imaging studies, human and other primate cell models including induced neurons from fibroblasts, and the generation of cognitively enhanced animal models will ultimately open a window into how to connect human specific gene patterns with human brain function.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.