In my last post I outlined a number of experimental studies using the Zebra Finch that have highlighted an additional dimension to the FoxP2 gene – not only is it upregulated in the avian brain throughout song development, but it is also downregulated in important song nuclei of adult birds in singing contexts that seem to involve ‘listening to one’s own song’ and subsequent error correction. Given that the pattern of expression of this gene is very similar in the developing brain of both humans and birds, one conclusion that has been drawn from this research is that FOXP2 downregulation may equivocally serve to facilitate online language processing function in the adult human brain.
General background on an intriguing new celebrity
Naturally, the next step has been to try and identify the downstream genes regulated by FOXP2 in order to build up a more detailed picture of how interactions between complex genetic networks influence key language-related disorders in humans. It is as a result of such efforts that another gene, although discovered almost a decade ago, has found its way into the spotlight: CNTNAP2.
In the developing human brain, CNTNAP2 is enriched in functionally specialised regions such as the frontal cortex, the stratium, and the dorsal thalamus (circuits within these regions are referred to as cortico-striato-thalmic circuits) central to executive function, planning and executing complex sequential movements, and thus potentially, language. This presents a striking contrast to the more uniform expression of Cntnap2 observed in the developing rodent brain where there is no evidence for enrichment in specific regions, suggesting a functional difference in the human version that could be related to vocal learning and modification.
Using a genomic screening technique which involved conducting expression analyses of human cortical fetal tissue in vivo, Vernes et al. (2008) uncovered a very interesting relationship between CNTNAP2 and FOXP2; they found that CNTNAP2 mRNA expression was lowest in cortical layers where FOXP2 levels were highest, concluding that the latter must modulate the former by binding to it and downregulating it dramatically.
Furthermore, a recent avian study by Panaitof et al. (2010), which again recruited the help of the Zebra Finch, reported significant sexual dimorphism in Cntnap2 expression in male and female Zebra Finch brains over the first 50 days post hatching, providing support for the prediction that Cntnap2 might be related to vocal learning and production (female Zebra Finches do not sing). They found that in early development (i.e. between15 and 25 days) before males undergo the first stage of song learning (the sensory-acquisition phase) at 30-35 days, Cntnap2 was equally expressed in the brains of both males and females. However, when song learning commenced this pattern changed, with Cntnap2 showing substantial enrichment and attenuation in important song circuits relative to surrounding tissue in male brains only.
The nature of CNTNAP2 as a downsteam target of FOXP2 and the neuroanatomical regions in which it is expressed in both humans and birds is concurrent with the notion that this gene is important for language development and performance in some way. Although the precise role of CNTNAP2 is not yet known, it is thought to contribute to several aspects of brain development and appears to encode a protein (CASPR2) integral to the shaping and functioning of neuronal and glial connections in the nervous system.
Effects of common CNTNAP2 variation on linguistic performance and processing
As Hannah explored in her overview of a recent paper by Dorothy Bishop, rare and common genetic variation in CNTNAP2 has been implicated in a number of neural disorders including Specific Language Impairment (SLI) and Autism Spectrum Disorders (ASDs). Characteristics of the latter often subsume but extend beyond language deficiencies to those of behaviour and social interaction.
In addition to their work on the relationship between FOXP2 and CNTNAP2, Vernes et al. (2008) reported the finding that within a sample of children with SLI, certain CNTNAP2 polymorphisms, or variants, were related to differences in performance on a nonsense word repetition task used to examine level of proficiency in phonological short-term memory. Although those with SLI commonly experience difficulties with this task, the differing levels of performance within this group and quantitative associations with CNTNAP2 polymorphisms suggests that variations of this gene cause differing levels of linguistic deficiency.
Key Study: Snijders (2010) More than words: Neural and genetic dynamics of syntactic unification
Wintz kindly signposted me to this paper (Snijders 2010), containing results which also suggest that common genetic variation in CNTNAP2 influences the way that we process language. Following a series of neuroscientific studies using fMRI and MEG techniques, Snijders and her colleagues claim to have identified key brain loci involved in online language processing and the ways in which these regions interact as our brains perform ‘syntactic unification’, that is, the computational process of arriving at the correct lexical-syntactic interpretation of a sentence by retrieving information about individual words from long-term memory and subsequently combining this information.
As this study is very extensive, for now I will focus on the findings of the MRI studies. Using this technique, the team were able to monitor the brain regions that were most active (areas in the left and right frontal lobes and the striatum) when normal, healthy subjects were presented with visual stimuli made up of word lists and sentences characterised by varying degrees of unification load (UL). In accordance with previous data suggesting that the right hemisphere plays a role in ‘broader’ and context-driven linguistic processing, right hemispheric regions (as well as left hemispheric regions) were implicated in processing sentences with high degrees of UL relative to sentences with low UL (reviewed in Jung-Beeman 2005).
The team then organised their subjects into two genotype groups based on a common CNTNAP2 variant, uncovering differences in the way that the members of each group used their brains in order to process the stimuli. Individuals with this variant potentially carry one or two copies of an allele which is associated with autism susceptibility. The participants were divided thus; individuals in one group carried two copies of the safe allele i.e. the ‘safe allele group’, where as individuals in the other group carried either one or two risk alleles i.e. the ‘risk allele group’.
Fascinatingly, although the overall brain network recruited during these tasks was similar across the board, the balance in activation patterns in frontal and temporal regions differed between ‘safe-allele’ and ‘risk allele’ groups. In addition, the ‘safe allele’ group showed higher right hemisphere activation when processing sentences with high UL relative to sentences with low UL and word lists than the ‘risk allele’ group, suggesting the latter group recruited a lexically-driven processing style, where as ‘safe allele’ individuals exhibited a more context-driven processing style. This also seemed to be reflected physically by anatomical brain differences between the two groups, with ‘risk allele’ individuals showing reduced grey matter in an active region of the right frontal lobe and less connectivity between active left hemispheric regions than the ‘safe allele’ individuals. Importantly, people with autism also rely on a significantly lexically-driven processing style which is probably why, for example, some have difficulty ‘getting’ jokes and comprehending metaphors (i.e. ambiguous language that requires a high unification load and depends on a high level of contextual inference).
CNTNAP2: Under the Influence – but still a strong character
The genotyping and neuroscientific studies outlined above provide compelling evidence to suggest that even CNTNAP2 variants within the normal range affect linguistic processing considerably, with some alleles of polymorphisms posing more of a threat than others. Given these influences, it is easy to imagine why the inaccurate regulation of CNTNAP2 and other downstream targets caused by mutations in FOXP2 would disrupt many aspects of linguistic performance. Interestingly, the CNTNAP2 variants identified in the above studies (Vernes et al. 2008; Snijders 2010) were not the same, and also seem to influence different aspects of linguistic performance. That ‘risk alleles’ on the CNTNAP2 polymorphism studied by Snijders (2010) causes a linguistic processing style similar to that demonstrated by those with autism but with no adverse effects suggests that this variant may have a magnifying effect if in the company of other genes that increase the risk of autism, but that it is not sufficient to produce the condition in absence of such genes. In short, these findings draw attention to the rich insight that can be gained from the study of variation within known language-related genes as well as the exciting endeavour of searching for new genes involved in language.
Jung-Beeman, M. (2005) Bilateral brain processes for comprehending natural language. Trends in Cognitive Sciences 9: 11, 513-518.
Panaitof, C. S. et al. (2010) Language-related Cntnap2 Gene is Differentially Expressed in Sexually Dimorphic Song Nuclei Essential for Vocal Learning in Songbirds. The Journal of Comparative Neurology 518: 1995-2018
Vernes, C. S. et al. (2008) A Functional Genetic Link between Distinct Developmental Language Disorders. The New England Journal of Medicine 359:22, 2337-2345.
Snijders, T. M. (2010) More than words: Neural and genetic dynamics of syntactic unification. RU Radboud Universiteit Nijmegen. PhD thesis.