Abstract
Two functional enhancers of FOXP2, a gene important for language development and evolution, exhibit several human-specific changes compared to extinct hominins that are located within the binding site for different transcription factors. Specifically, Neanderthals and Denisovans bear the ancestral allele in one position within the binding site for SMARCC1, involved in brain development and vitamin D metabolism. This change might have resulted in a different pattern of FOXP2 expression in our species compared to extinct hominins.
Main text
Mutations in the coding region of FOXP2 are known to cause speech and language impairment (Watkins et al. 2002; Shriberg et al. 2006; Reuter et al. 2017). Modern humans and extinct hominins (Neanderthals and Denisovans) are thought to be endowed with distinctive language abilities, but no differences in the coding region of FOXP2 have been found between species (Benítez-Burraco et al. 2008; Johansson 2015). Changes in the expression pattern/level of FOXP2 have been thus hypothesised to contribute to the species-specific idiosynchratic language profile. Interestingly, Neanderthals bear the ancestral allele of a binding site for the transcription factor POU3F2 within intron 8 of FOXP2, which is more efficient in activating the transcription of the target gene (Maricic et al. 2013). Accordingly, higher levels of FOXP2 might be expected for this hominin species. In humans, partial duplications of FOXP2 are related to delayed speech and language development (nssv13656147) and autism features with intellectual disability (nssv13648704).
We have recently uncovered two enhancers of FOXP2 in the intergenic region between FOXP2 and its adjacent MDFIC gene and validated their functionality by CRISPR/Cas9 (Torres-Ruiz et al. 2016). Deletion of any of them in the SK-N-MC neuroblastoma cell line downregulates FOXP2 and decreases FOXP2 protein levels. This suggests that both elements upregulate FOXP2 in vivo in brain cells. In this letter, we describe several differences in the sequences of these two enhancers between Neanderthals, Denisovans, and modern humans, which support the view that the expression of FOXP2 may be differentially regulated in these three hominin species, plausibly impacting on the sensorimotor loops for auditory-vocal control to which FOXP2 contributes.
In doing so, we first compared the sequence of the two enhancers in several species of interest, namely, vertebrates in which imitative vocalization has been extensively studied in the context of language evolution. These include songbirds (zebra finch, collared flycatcher, white-throated sparrow, medium ground finch), cetaceans (dolphins, killer whales), bats (black flying-foxes, megabats, David’s myotis, microbats, big brown bats), primates (humans, chimps, gorillas, orangutans, gibbons, rhesus, baboons, marmosets, bushbabies), and rodents (mice). We then aligned the Denisovan and Neanderthal sequences with the human reference genome (GRCh37), looking for differences in the sequences of the two enhancers. Finally, we interrogated in silico whether the observed differences affect the regulatory properties of the two enhancers, and ultimately, FOXP2 expression.
We found that these two enhancers are absent in songbirds (Figure 1). The proximal one (FOXP2-Eproximal) is found in primates, bats and cetaceans, but not in rodents, whereas the distal enhancer (FOXP2-Edistal) is partially conserved in cetaceans, some bats, and most primates (Figure 1). Interestingly, most of the predicted binding sites for transcription factors within the two enhancers are absent in the primate species that are most distantly related to humans (bushbabies and marmosets) (Figure 1). Regarding extinct hominins, in spite of the high degree of homology with modern humans, we found some differences between the human reference genome (GRCh37) and the genomic sequences from Neanderthals and Denisovans. The Denisovan FOXP2-Eproximal exhibits six differences with the modern human sequence, with two of these changes (highlighted in blue in Figure 1) located within predicted targets for transcription factors. The Neanderthal FOXP2-Eproximal shows five differences with the human enhancer, with two of them (highlighted in red in Figure 1) located within the predicted binding sites for the transcription factors POLR2A and SMARCC1. Only one change is shared by both species (114460158T>G, hg19), which affects the binding site for SMARCC1. Regarding FOXP2-Edistal, there are no differences between the Denisovan sequence and the human sequence, whereas we found two differences with the Neanderthal enhancer, although they are located outside any predicted binding site for transcription factors (Figure 1).
FOXP2 is a gene important for speech and language development and evolution (Scharff and Petri 2011; Graham and Fisher 2015). In view of the identity observed at the protein level between the three hominin species, we regard this nucleotide difference in the binding site for SMARCC1 within FOXP2-Eproximal of interest for inferring evolutionary modifications of FOXP2 expression in the human clade. As noted, both abnormally low and abnormally high levels of FOXP2 protein result in abnormal cognitive phenotypes, which entail problems with speech and language. SMARCC1, a component of the large ATP-dependent chromatin remodelling complex SNF/SWI, plays an important role in the development of the forebrain, particularly, in neurogenesis (Narayanan et al. 2015). The implicated regulatory mechanism also involves PAX6 (Ninkovic et al. 2013), which controls FOXP2 expression (Coutinho et al. 2011). It has been hypothesised that selected differences in the coding and/or regulatory regions of PAX6 and some of its partners might account for some of the cognitive differences between Neanderthals and humans (see Benítez-Burraco and Boeckx, 2015 for discussion). Interestingly, one of these partners is SPAG5, which helps PAX6 to regulate neuronal proliferation (Asami et al. 2011) and which exhibits several fixed changes in modern humans (Green et al. 2010; Meyer et al. 2012; Prüfer et al. 2013).
SMARCC1 is also involved in vitamin D-coupled transcription regulation (Hawes et al. 2015). Vitamin D deficiency reduces the amount of Foxp2-expressing cells in the developing cortex (Hawes et al. 2015). Interestingly too, core candidates for the evolution of our ability to learn and use languages (aka, our language-readiness) are related to vitamin D homeostasis and function (see Benítez-Burraco and Boeckx 2015 for details). At the same time, Western European Neanderthals have been claimed to suffer from vitamin D deficit, which might be related to some of their distinctive cognitive features (see Greenfied 2015 for details). Also, in present-day human populations, low vitamin D levels seem to correlate with the severity of the symptoms of cognitive diseases entailing language deficits, like autism-spectrum disorder (Bakare et al. 2011; Jia et al. 2015) and schizophrenia (Amato et al. 2010; Yüksel et al. 2014). Intriguingly, the expression of SPAG5, highlighted above, is regulated by VDR, the receptor of vitamin D.
We expect that this analysis in silico helps achieve a better understanding of the evolutionary changes in the regulatory landscape of FOXP2 that seemingly contributed to refine the biological machinery supporting human speech and language. A second step will be to provide with a functional validation of our results, by mimicking the Neanderthal/Denisovan change in the binding site for SMARCC1 in the same human neuroblastoma cell line that we used for testing the functionality of the two human enhancers.
Methods
The identification, characterization, and functional validation of the two FOXP2 enhancers were performed as described in Torres-Ruiz et al. (2016). For the alignment of the vertebrate sequences, we used the UCSC/Penn State Bioinformatics comparative genomics alignment pipeline. The high-coverage Neanderthal and Denisovan sequences of the two enhancers were retrieved from The Neanderthal Genome Project (http://cdna.eva.mpg.de/neandertal/altai/AltaiNeandertal/bam/) and The Denisova Genome Consortium (http://cdna.eva.mpg.de/denisova/), respectively. The sequences were aligned with Clustal Omega to find the nucleotide changes. For determining the regulatory properties of the enhancers, we used the Encyclopedia of DNA Elements (ENCODE, https://genome.ucsc.edu/ENCODE/), searching for distinctive DNA hallmarks (DNase I hypersensitive sites, presence of histones with specific post-translational modifications, in particular histone H3, lysine 4 monomethylation (H3K4Me1) and H3 lysine 27 acetylation (H3K27Ac)), and particularly, for target sequences for potential co-activators and co-repressors, as revealed by chromatin immunoprecipitation followed by deep sequencing (ChIP-seq).