deficiency causes a complex speech and language disorder
The KE family comprised of three generations in which severe speech and language impairment (developmental verbal dyspraxia, DVD) segregated with an autosomal dominant mutation in the transcription factor FOXP2. The mutation is an arginine-to-histidine substitution (R553H) located in the forkhead DNA-binding domain, which disrupts DNA-binding and transactivation properties (Lai et al., 2001; Vernes et al., 2006; Mizutani et al., 2007), (Fig.1).
Although the monogenetic inheritance pattern of the disorder was evident at the time of its first description (Hurst et al., 1990), its nature and clinical diagnosis became a matter of polemics (Fletcher et al., 1990; Gopnik, 1990b), as the manifestation of the dysfunction and the exact language defect appeared to be rather complex.
Linguist M. Gopnik defined the disorder as a dysphasia – i.e. disturbed ability to communicate – due to inability of patients to infer general syntactic rules, like the generation of plurals, tenses, genders. This lack of general grammar rules was especially evident in spontaneous speech, writing or grammatical judgement; however, despite disturbed ability in usage of grammatical constructs, the affected individuals seemed to be able to assign their meaning correctly. Gopnik suggested that the defect was not a consequence of a general cognitive problem, but was specific to grammar domain ‘because the language skills that are not impaired are at least as complex as those which are’ (Gopnik, 1990a).
Investigations of the KE family by Vargha-Khadem et al. concluded that the most prominent phenotype of the disorder is a deficit in the language production-processing system rather than in grammar-processing or morphosyntactic rule formation. The disorder was defined as speech dyspraxia (from ‘praxis’- greek for process, activity), with a central defect in oro-facial motor movements; the extent of the deficit is such that a speech of some affected individuals is unintelligible for the naïve listener. These articulation problems are reflected in a relative immobility of the lower face and mouth and consequently affect oral and facial movements when motor actions are performed simultaneously and sequentially (Vargha-Khadem et al., 1995). Accordingly the orofacial dyspraxia can be robustly detected using three tests: word repetition, non-word (i.e. meaningless words) repetition, and simultaneous and sequential oro-facial movements – all three essentially representing a measure of orofacial motor coordination (Vargha-Khadem et al., 1998; Watkins et al., 2002a). The motor phenotypes were instrumental to the precise assignment of affected and unaffected family members for genetic mapping and identification of the underlying FOXP2 mutation. The final confirmation of the causative genomic region came from an individual with a chromosomal translocation unrelated to the family (Lai et al., 2001).
However, intellectual and linguistic deficits are considered part of the condition; examination of affected KE family members using a battery of linguistic tests (13 tasks examining general language functions and 4 specifically addressing grammar) showed impairments in all but one: object naming. Impairments in language tasks included word and sentence repetitions, lexical decision (to define if the word is real), phoneme deletion and addition from words and non-words, non-word spelling and reading, rhyme production; grammatical tests addressed the ability to process inflections (i.e. knowledge of morphosyntactic rules), tense production and receptive knowledge of sentence embedding in the form of relative clauses (Vargha-Khadem et al., 2005).
Moreover, in addition to the pronounced language problems, the authors reported lower IQ scores in the KE family, and concluded that cognitive impairments are not confined to morphosyntax, but rather extend to verbal and non-verbal domain generally (Vargha-Khadem et al., 1995). Permanent lack of appropriate communication endured from early childhood might contribute to a slightly lowered IQ, a secondary effect to DVD. This idea is supported by Watkins et al. (2002a) stating that ‘a disorder may adversely affect the development of intelligence or of the skills required to maintain a given level of intelligence’ Given the potential presence of a cognitive component in the KE family disorder, it was relevant to dissociate it from the motor-articulation deficits, and to identify which brain structures were affected (Liegeois et al., 2003). Therefore a covert verb generation task was used: the participants were asked to generate the verb to a given noun in their thoughts, without pronouncing it or performing any movement. While performing this task under functional magnetic resonance imaging (fMRI) monitoring, affected KE family members showed underactivation of Broca’s area (classically associated with syntactic aspects of speech) and in Brodmann’s area (semantic aspects), together with atypical patterns of overall cortical activation: higher and diffused activation of the right hemisphere as compared to the left, and posterior compared to anterior (Fig.1). In this task underactivation was also detected in striatal structures (right putamen/globus pallidus), interpreted as reflecting articulation planning impairments. Interestingly, the overt (spoken) verb tasks – generation and repetition – yielded largely similar misactivation patterns, suggesting that at least in these tests the phenotype largely relied on upstream cortical and striatal abnormalities (Liegeois et al., 2003). Consistently with fMRI observations, grey matter volume measurements in structural imaging studies revealed alterations in motor related structures with the most striking effects in striatal areas (caudate nucleus, putamen and globus pallidus), but also in cerebellum (ventral and posterior lobe); several cortical areas were affected morphologically as well: Brodmann’s and Broca’s areas, cingulate, sensorimotor, motor, posterior temporal, anterior insular, medial occipitoparietal and inferior frontal cortices (Vargha-Khadem et al., 1998; Watkins et al., 2002b; Belton et al., 2003). Comparison of affected KE-family members to patients with acquired Broca’s aphasia (i.e. a collection of language disorders caused by damage to the brain, in this case a left hemisphere stroke) was employed to partially discriminate cortical-driven and striatal aspects of the performance: the difference between the two conditions was detected only within the tests of verbal fluency and lexical decisions, but surprisingly did not seem to lie within oral praxis or receptive and expressive language deficits (Watkins et al., 2002a).
Ackermann (2014) discussed fundamental differences in the profiles of speech motor deficits in verbal dyspraxia associated with FOXP2 mutations versus Parkinsonian motor speech disorder, caused by basal ganglia dysfunctions. Acquired impairments in speech performance due to striatal damage are characterized by execution of orofacial movements with ‘undershooting’ gestures and disruptions of prosodic aspects of verbal utterances (i.e. rhythm, stress, and intonation), while communication disorders due to fronto-opercular cortex or anterior insula damages in the language-dominant hemisphere resemble FOXP2 phenotypes to much greater extent (Ackermann et al., 2014).
Hence, although it is difficult to conclude which language deficit in FOXP2 deficient patients arises primarily, misarticulation or linguistic, at least in adults the primary problem may lie upstream of the motor system, and has a significant contribution from cortical dysfunctions.
Characterization of the KE family was followed by reports of other patients showing similar phenotypes with FOXP2 loss-of-function mutations and disruptions of genomic regions residing in the vicinity of the gene (O’Brien et al., 2003; MacDermot et al., 2005; Lennon et al., 2007; Tomblin et al., 2009; Palka et al., 2012).
Disruptions of FOXP2 chromosomal region (7q31) have been linked to a spectrum of neurodevelopmental disorders, such as autism and schizophrenia (as well as major depression, dyslexia, FTLD, ADHD); these pathologies exhibit some level of association with FOXP2 polymorphisms, often in parallel with the expressivity of language endophenotypes (Gong et al., 2004; Li et al., 2005; Sanjuan et al., 2005; Sanjuan et al., 2006; Padovani et al., 2010; Tolosa et al., 2010; Schaaf et al., 2011; Spaniel et al., 2011; Ribases et al., 2012; Wilcke et al., 2012; Li et al., 2013; Corominas et al., 2014; McCarthy-Jones et al., 2014). However, data on the direct genetic association between FOXP2 and many of the aforementioned disorders remain controversial (Newbury et al., 2002; Gauthier et al., 2003; Laroche et al., 2008).
A molecular link between FOXP2 and mental disorders has been established in functional genomics studies through a number of genes and pathways involved in a several distinct mental pathologies (Spiteri et al., 2007; Vernes et al., 2007; Konopka et al., 2009; Vernes et al., 2011). Specifically, among the best described genes directly regulated by FOXP2 are:
CNTNAP2 – a gene substantially enriched in frontal gray matter and associated with autism, Tourette’s syndrome and severe recessive disorder involving cortical dysplasia and focal epilepsy (Vernes et al., 2008)
MET – component gene involved in human temporal lobe development, associated with autism (Mukamel et al., 2011)
DISC1 – a leading candidate susceptibility gene for schizophrenia, bipolar disorder and recurrent major depression, which has been implicated in other psychiatric illnesses of neurodevelopmental origin, including autism (Walker et al., 2012).
SPRX2 – a gene responsible for speech dyspraxia and mental retardation which accompany a form of sylvian epilepsy (Roll et al., 2006), which independently links SPRX2 to speech and language
uPAR – an interaction partner of SPRX2 which is associated with autism spectrum disorder (ASD) (Eagleson et al., 2010; Roll et al., 2010).
FOXP2 may be linked to autism and intellectual disability via interacting partners such as FOXP1 (Bacon and Rappold, 2012) and TBR1. The latter is a recently described autism candidate gene (Deriziotis et al., 2014) and protein-interaction studies of ASD susceptibility genes have shown a direct interaction with FOXP2(Sakai et al., 2011; Corominas et al., 2014).
Thus, several lines of evidence strongly suggest that FOXP2 is part of larger molecular networks underlying distinct language and social cognitive dysfunctions To summarize, human studies on FOXP2 gene function suggest that it is involved in complex motor phenotypes of speech and language processing. In addition, FOXP2 role in both semantic and syntactic aspects of language is supported by pronounced defects in grammar rules formation and problems in lexical decision-making, in agreement with morphological abnormalities in the cortex including Broca’s and Brodmann’s area. Evidence for FOXP2 role in phonological domain of language derives from recent studies on healthy individuals. FOXP2 polymorphisms were shown to contribute to the normal inter-individual variability in hemispheric asymmetries and frontal cortex activation patterns for speech and language audio (dichotic listening task), and also visual (reading tasks) perception (Pinel et al., 2012; Ocklenburg et al., 2013). Cortical lateralization impairments, characteristic of FOXP2 dysfunction, are well described in language and literacy per se and often diagnosed in autism and speech and language disorders (SLI) (De Fosse et al., 2004; Bishop, 2013). Co-morbidity of FOXP2-based language and molecular phenotypes with a variety of mental disorders, ASD in particular, hints to FOXP2 involvement in more complex cognitive functions than simple processing of certain aspects of language. In this respect Corballis (2004) suggested a link between FOXP2 and the mirror neuron system, which underlies many critical symptoms of autism. ‘Thus, while studies of FOXP2 can offer insights into relevant neural pathways, it is not a ‘gene for speech’.’- Fisher (2007).
The discovery of FOXP2 as the first gene involved in speech and language development led to studies on its role in human evolution. Using comparative analysis of human, chimpanzee, orangutan and mouse DNA sequences, two independent groups established that FOXP2 is a remarkably conserved gene that was the object a selective sweep in the human lineage. Positive selection in the coding sequence is thought to have converged on two amino acids within exon 7 that distinguish humans from other primates, while only three amino acids substitutions differentiate mouse Foxp2 from humans (Enard et al., 2002; Zhang et al., 2002) (Fig. 2). Further investigations suggested that selection pressure may have been exerted beyond the two human specific substitutions of the coding FOXP2 region, on sites located near the gene (Ptak et al., 2009; Maricic et al., 2013). Genome-wide scans for positively selected genes during mammalian evolution identified only 5 transcription factors in the hominid branch with a selection rate comparable to FOXP2 and none of them was associated with brain function, but with spermatogenesis and the immune system (Kosiol et al., 2008; Enard, 2011).Enard proposed that a murine model, carrying two ‘human’ amino acid substitutions within Foxp2, could provide valuable information of the evolution of mammal circuits underlying speech and language in humans (Enard, 2011). Therefore, he and his colleagues have introduced the two human specific substitutions in the orthologous mouse locus, generating a homozygous knock-in Foxp2 ‘humanized’ mouse (Foxp2-Hum) (Table 1) (Enard et al., 2009). Although Foxp2 is expressed in vitally important organs such as heart, lung and guts (Shu et al., 2001), the exhaustive screen for almost 300 different phenotypic parameters did not produce any evidence for effects of the are mapped on the phylogeny of FoxP2 sequences from vertebrates. The bars on the right site depict the ratio of amino acid changes to the length of the terminal branches. Asterics indicate species with evidence for vocal learning. B, Amino acid changes in the tree are plotted for each position of the human FOXP2 protein sequence. ZnF, zinc finger; LZ, leucine zipper; FOX, forkhead-box domains, the two human amino acid changes are shown as red lines. Adapted from Enard (2011).
Foxp2Hum allele in any organ system except the central nervous system. Here, anomalies were revealed in behavioral, electrophysiological, histological and molecular parameters within the brain. Furthermore, such anomalies were most prominent within cortico-striatal circuits (Enard et al., 2009; Reimers-Kipping et al., 2011; Schreiweis et al., 2014).
These findings are in agreement with the cortico-striatal core of phenotypes in affected KE-family members. The authors suggested that human-specific mutations in Foxp2 in mice could model aspects of speech and language evolution in humans (Enard et al., 2009).
The neural circuits underlying uniquely human functions, involving abstract thinking and the cognitive mechanisms of language processing (as, for example, syntax), are located in the cortex. This brain structure expanded significantly in the human lineage relative to other primates, developing new regions in the frontal and parieto-temporal lobes. Cortical volume expansion is mechanistically based on increases in interecellular space and cell number, due to higher cortical neurogenesis, enhanced elaboration of dendritic trees and expansion of neuropil areas, especially in the prefrontal cortex (Geschwind and Rakic, 2013).
As discussed below (in Foxp2 cellular functions, Foxp2 in the mouse cortex), FoxP2 is strongly engaged in processes of neurite outgrowth and cortical neurogenesis (Spiteri et al., 2007; Vernes et al., 2007; Vernes et al., 2011; Tsui et al., 2013; Chiu et al., 2014); in this light it is interesting to examine how FOXP2 role in cortical evolution is gradually emerging from comparative genomic studies: When addressing evolution of the coding genome at the single nucleotide level it is accepted that brain-expressed genes are not specific for positive selection in human lineage (Wang et al., 2007; Geschwind and Rakic, 2013). However, FOXP2 appears to be one of the salient exceptions, as inferred from Enard’s studies described above, as well as from the analysis of Neanderthal and Denisovan genomes sharing the human-derived form of FoxP2 right before the lineages split (Krause et al., 2007; Meyer et al., 2012). Many human brain-specific evolutionary accelerated changes lie within non-coding nucleotide stretches, i.e. enhancers and promoters (Capra et al., 2013). Visel et al. in their comparative study of conserved non coding regions relevant to telencephalic evolution, found a dozen of elements in close and distant proximity of FOXP2, as well as many more close to the genes constituting FOXP2 molecular network. Among these is the highly conserved transcription factor POU3F2, which has been suggested as candidate for having caused a recent selective sweep of FOXP2 in the human lineage (Maricic et al., 2013; Visel et al., 2013). In a separate study using comparative epigenetic profiling of human, mouse and macaque corticogenesis to identify enhancers which gained activity in human evolution, FOXP2, although not identified as one of the hub genes, showed substantial connectivity to gain-enriched modules of co-expressed genes(Reilly et al., 2015).
Konopka et al. performed comparative analyses of transcriptomes from the telencephalon of human, chimpanzee and macaque telencephalon. They demonstrated a striking increase in transcriptional complexity specific to the human lineage in the frontal lobe, while genes expression in the caudate nucleus was conserved. In particular, the human prefrontal cortex is enriched in alternatively spliced genes involved in neuron projections, neurotransmitter transport, synapses, axons, and dendrites, as well as genes implicated in schizophrenia. Among the most differentially connected genes, this module contained FOXP2, and 13 genes that overlap with previously identified FOXP2 targets, as well as its dimerization partner – FOXP1 (Konopka et al., 2012).
Taken together comparative genomic data support the hypothesis that FOXP2 was actively recruited in the formation of the most evolved structures of the human cortex during hominin evolution.
If the human telencephalon is a structure, providing unique capacities, how could we approach its origins? Insights might come from the application of the concept of ‘deep homology’, stating that new structures do not arise de novo. Instead there is a continuum of events which follows an evolutionary constraint imposed by pre-existing genetic regulatory circuits, which were initially established in early metazoans. Classical examples of deep homology are the re-appearance of eyes in widely divergent organisms, based on similar genetic components such as ciliary phototransduction and melanogenic pathways, or tetrapod limbs and fish fins which are both based on deeper homology in the network of Hox genes (Shubin et al., 2009). When evolutionary constraints are further restrained by the structure, composition and dynamics of the developmental system (the fusion of these two fields is also called ‘evo-devo’ concept) there is even less room for deep phenotypic variability. The concepts of ‘deep homology’ and ‘evo-devo’ are classically applied to evolution of structures and forms, but could be adapted to neural circuits and behaviours (Robinson et al., 2008; Scharff and Petri, 2011). In the case of FoxP2, genomic sequence, developmental expression pattern and molecular network are conserved in distant mammals (such as human and mice) (Lai et al., 2003)(Fig.3). Within this framework mouse Foxp2 could serve as an entry point to study ancestral substrates underpinning complex aspects of FOXP2-dependent human cognition.
Some evidence exists for speech and language modulation by DA in humans. Neurological and psychiatric studies indicate that dopamine receptor antagonists disrupt vocal motor control and lead to the development of uncontrolled laryngeal spasms. Voice, articulation, phonological processing and syntactic complexity deterioration in Parkinsons disease patients – a disorder of dopaminergic neurons loss – is the most striking example of DA-dependent voice and speech problems. Similarly, alterations in DA levels or in nigrostriatal dopamine release are linked to speech related defects as diverse as vocal tics in Tourette’s syndrome, auditory hallucinations in schizophrenia and stuttering (Simonyan et al., 2012).
The identification of FOXP2 transcriptional targets has provided valuable insights into regulated cellular processes. High-throughput approaches such as FoxP2 chromatin-immunoprecipitation, coupled with gene expression profiling in cell cultures transfected with FoxP2 variants, or mouse and human embryonic material have been used (Spiteri et al., 2007; Vernes et al., 2007; Konopka et al., 2009). Differentially expressed direct and indirect targets have been functionally annotated using ontology categories defining specific biological processes. Gene Ontology (GO) categories characteristic for developmental FOXP2 targets in humans (Fig. 6) largely overlap with mouse GO categories. Detailed analysis of mouse data complements with the categories of neurogenesis, neuron projection morphogenesis, cell localisation functions as well as synaptic plasticity and spine formation (Vernes et al., 2011). These functions are supported by studies of Foxp2 mutant mice in vivo with pronounced affects on development and function of cortico-striatal circuits (Groszer et al., 2008; Enard et al., 2009; Schulz et al., 2010; Chiu et al., 2014).
Foxp2 role in cortical neurogenesis is well described in mice, where Foxp2 ectopic misexpression in cortical progenitors impairs in vivo radial migration, neurite maturation and synaptogenesis (Clovis et al., 2012; Sia et al., 2013), and affects progenitors type transitions during neuronal differentiation (Tsui et al., 2013; Chiu et al., 2014)(Fig.8). FoxP2 specific role in neurogenesis is further supported by studies in chicken and zebra finch (Rousso et al., 2012; Thompson et al., 2013).
In zebra finches, the major functional modules associated with singing and regulated by FoxP2 relate to synaptic plasticity (more specifically, the suppression of postsynaptic plasticity) via FoxP2 interconnection with genes linked to MAPKK and NMDA receptors function, actin cytoskeleton regulation, and tyrosine phosphatase (Hilliard et al., 2012).
Cocaine stimulation of heterozygous mice (Foxp2-R552H/+-Enu, Table1) revealed neuronal activity alterations in NuAc mediated by genes involved in calcium signaling, among which voltage-dependent calcium channel alpha 1G (Cacna1g) has been genetically associated with ASD, providing further support for the role of Foxp2 in social behavior (Mombereau et al., submitted).
Mouse models in Foxp2 research: motor learning and ultrasound vocalizations
Foxp2 expression pattern is highly conserved across four species of muroid rodents: two species of singing mice (Scotinomys teguina and S. xerampelinus), their close relative the deer mouse (Peromyscus maniculatus), and more distantly related laboratory mouse Mus musculus (Campbell et al., 2009).
Foxp2 distribution in the mouse brain is rather broad and extends to lower cortical layers, striatum, mesolimbic and nigrostriatal dopaminergic systems, thalamic somatosensory areas, Purkinje cells of cerebellum and inferior olivary complex, olfactory system and ascending auditory and visual relays (Fig. 7) (Ferland et al., 2003; Campbell et al., 2009). This expression pattern suggests that Foxp2 is unlikely to regulate circuits selectively governing verbal or vocal functions, but rather specifically interacts with genetic factors that define relevant circuits and pathways. The relevant functional circuitry in adult mice has been suggested to lie within networks controlling fine motor output, multimodal sensory processing and sensorimotor integration (Campbell et al., 2009).
A concise overview of the existent genetically modified Foxp2 mouse models and their basic phenotypes has been provided in a recent review by French and Fisher (2014), Table 1.
Because of the known relevance of cortico-striatal motor circuitry for auditory guided vocal communication (Wohlgemuth et al., 2014), supported by high conservation of striatal circuitries among vertebrates, most attempts to experimentally access Foxp2 functions in mice have used motor learning paradigms. The heterozygous loss-of-function model, bearing KE family etiological mutation: Foxp2-R552H/+-Enu (Table 1), manifests motor learning impairments in three distinct assays: voluntary-controlled running wheel, automatically accelerated rotarod and auditory-motor associations learning (accessing auditory-motor integration impairments potentially relevant for speech acquisition); the latter approach also revealed a worse learning curve in another heterozygous model – Foxp2-S321X/+. Striatal electrophysiological abnormalities were observed in Foxp2-R552H/+ mice, including reduced cortico-striatal long-term depression (LTD) in brain slices, and abnormally high ongoing striatal activity along with dramatic alterations of striatal plasticity during motor skill acquisition in vivo (Groszer et al., 2008; French et al., 2012; Kurt et al., 2012).
Foxp2 humanized mice (Foxp2-Hum) a model which effectively manifests gain-of-function profile, show enhanced striatal LTD compared to WT and Foxp2 heterozygote mice (Enard et al., 2009; Reimers-Kipping et al., 2011). Furthermore, humanized mice demonstrate faster proceduralization of action sequences in conditional T-maze paradigm with spatial cues. This learning paradigm accesses declarative (place-based) learning transitions to procedural (response-based) learning, and reveals the speed of information transfer from dorso-medial to dorso-lateral striatum (Schreiweis et al., 2014).
Table of contents :
FOXP2 deficiency causes a complex speech and language disorder
Foxp2 as an entry point to study molecular and neural networks contributing to cognitive aspects of speech and language
The study of animal models provides insights into conserved FoxP2 functions
The role of FoxP2 in development
Activity dependent function of FoxP2 in mature brain: evidence for a role in social behaviour and vocalizations
FoxP2 cellular functions
Mouse models in Foxp2 research: motor learning and ultrasound vocalizations
Foxp2 in the mouse cortex
Materials and Methods
Analysis of projections: brain stereotaxic injections
Ultrasonic vocalizations (USV)
Cell type–specific mRNA purification by translating ribosome affinity purification (BacTRAP)
1. Generation and characterization of Foxp2 cortex-specific homozygous knockout mice
1.1. Cortical Foxp2 ablation does not affect gross cortical morphology
1.2. Foxp2 cKO animals do not show gross projection abnormalities
1.3. Postnatal development of Foxp2 cKO mice and WT littermates is indistinguishable
1.4. The role of cortical Foxp2 in DA signaling related behavior
1.5. Social interaction defects in Foxp2 cortical knockout mice
1.6. The role of cortical Foxp2 in modulating ultrasonic vocalizations (USVs)
2. Molecular profiling of lower cortical neurons in Foxp2+/- mice 82
3. Autism related gene-Mint2- is downregulated in the cortex of Foxp2 cKO mice